Audio Watermarking Different Algorithms and its Applications

Audio Watermarking Different Algorithms and its applications

2014

By Prof. Vishnu Narayan Saxena & PROF. POOJA SAXENA

ABOUT THE BOOK

Security has always been a major concern for human from very ancient age to this new age of science. The aspect and meaning of security is kept on changing from day by day as the science and technology changes. Now a day’s internet has been converted into a very big market place where everything is available for buying and selling again we can’t ignore the importance of an internet for the promotion of our ideas or our product. How we can protect our Data from an unauthorized copying and distribution over internet this is the main driving force for the development of a technology called watermarking. Watermarking is a technology in which we add some watermark(for example data ,picture, any random numbers) with the original host signal(Text, Audio, Video) in such a way that it will not affect the quality of an original signal as well as the added watermark should be robust against various intentional or unintentional signal processing attacks. In this book we describe about watermarking in very systematic and in very easy way Its various applications, Requirements , various attacks on watermarking ,Different domain of watermarking and various algorithms developed

for embedding the watermark information with the host signal. Prof Pooja Saxena & Prof. Vishnu Narayan Saxena

ABOUT THE AUTHORProf Vishnu Narayan Saxena is currently working with U.P.C. an AICTE approved Institution under the department of technical education (D.T.E). Government of Madhya Pradesh India. He has received his Master of engineering in communication control and networking from M.I.T.S. Gwalior and his Bachelor of engineering from I.T.M. Gwalior. Prof Saxena has been in field of Technical education and research for last 11 year. He has been published many International/national research papers on different field of an engineering and has been published

many e-books with different repudiated publications. His many articles have been published and available online at Scribed, Lulu publication, Slide share, Amazon. He has been participated in many FDP organized by different IITs (Indian institute of technology) and NITs (National institute of technology) and completed many online course (ALISON).His field of research and interest are Digital signal processing, Digital Image Processing, Audio signal processing, Wavelets and filter design.

DEDICATIONI want to dedicated this work to my father Mr B.N. Saxena by whom hard working capacity and consistency has been transferred into me. My wife

Pooja who always appreciate me to do good job and for ignoring all of my mistakes committed by me knowingly or unknowingly and My cute daughter Poorvi who developed into me a sense of responsibility and patience. I am very thankful to my wife and co author of this book prof. pooja saxena who discussed to me about every topic of this book and help me to improve our work she always brings to me with new ideas and suggestions. I am very grateful for every person by whom I could learn. I am also grateful for all the situation and incidences by which I learn something Again I want to say thanks technology (Google search engine and internet) which help me a lot to understand this subject and help me to share my views to others. I am also thankful to all the authors and writers who share their research work with us.

CHAPTER- 1 INTRODUCTION1.1 Definition of audio watermarking “Digital audio watermarking is a technique for embedding additional data (for example image) along with audio signal. Embedded data is used for copyright owner identification”

“Audio watermarking is used to hide information in audio signal. This data should not be hearable to human ear and it should robust, so that it could be used for the purpose of Intellectual Property Rights. ” The watermark is a signal embedded into the host media to be protected, such as an image or audio or video. It contains useful certifiable information for the owner of the host media, such as producer's name, company logo, etc. the watermark can be detected or extracted later to make an assertion about the host media. Audio watermarking is a technique that hides copyright information into the digital audio signal. Embedded data not only must be imperceptible but also should resist attacks and

other types of distortions trying to remove or neutralize the watermark picture. Audio watermarking is used to hide information in audio signal. This data should not be hearable to human ear and it should robust, so that it could be used for the purpose of Intellectual Property Rights. In the process of watermarking there are two signals [1] Host signal: The original signal (for example audio, video, image, text) which is to be protected from unauthorized copying and distribution.[2]Watermark: Watermark may be a binary data sequence or an image which is to be embedded as a proof of ownership with host signal.

Fig (1.1) Process of watermarking

1.2 Need of watermarking With the growth of the Internet, unauthorized copying and distribution of digital media has never been easier. As a result, the music industry claims a multibillion dollar annual revenue loss due to piracy. Normally an application is developed by a person or a small

group of people and used by many. Hackers are the people who tend to change the original application by modifying it or use the same application to make profits without giving credit to the owner. So we do require a technique which can protect our data from unauthorized copying and distribution and can provide copyright owner identification for our digital data over internet. Digital watermarking technology is now drawing attention as a new method of protecting unauthorized copying of digital content. A digital watermark is an imperceptible signal added to digital multimedia data (namely, audio, video, or image), which should remain even after several signal processes or potential attacks.

1.3 Types of watermarking According to the type host signal(1)Text watermarking: If host or original signal is text then it is known as text watermarking.(2)Audio watermarking: If host or original signal is an audio signal then it is known as audio watermarking.(3) Video watermarking: If host or original signal is video then it is known as video watermarking. According to the detection process(1)Blind watermarking: if original signal is not required for the extraction of watermark then it is known as blind watermarking. Blind watermarking is also known as public watermarking.

(2)Non blind watermarking: if original signal is required for the extraction of watermark then watermarking is known as non blind watermarking. Non blind watermarking is also known as private watermarking

1.4 Applications of watermarking

[I] Ownership protection and proof of ownership: In ownership protection application, the watermark embedded contains a unique proof of ownership. The embedded information is robust and secure against attacks and can be demonstrated in a case of dispute of ownership. There can be the situations where some other person modifies the embedded watermark and claims that it is his own. In such cases the actual owner can use the watermark to show the actual proof of ownership.

[II]Authentication and tampering detection: In this application additional secondary information is embedded in the host signal and can be used to check if the host signal is tampered. This situation is important because it is necessary to know about the tampering caused to the media signal. The tampering is sometime a cause of forging of the watermark which has to be avoided .In the content authentication applications, a set of secondary data is embedded in the host multimedia signal

and is later used to determine whether the host signal was tampered.

[III]Copy control and access control In the copy control application, the embedded watermark represents a certain copy control or access control policy. A watermark detector is usually integrated in a recording or playback system, like in the proposed DVD copy control algorithm or during the development Secure Digital Music Initiative (SDMI). After a watermark has been detected and content decoded, the copy control or access control policy is enforced by directing particular hardware or software operations such as enabling or disabling the record module. These applications require watermarking algorithms resistant against intentional attacks and signal processing modifications, able to perform a blind watermark detection and capable of embedding a non-trivial number of bits in the host signal. [IV]Information carrier The embedded watermark in this application is expected to have a high capacity and to be detected and decoded using a blind detection algorithm. While the robustness against intentional attack is not required, a certain degree of robustness against common processing like MPEG compression may be desired. A public watermark embedded into the host multimedia might be used as the link to external databases that contain certain additional

information about the multimedia file itself, such as copyright information and licensing conditions

[V]Finger printing Additional data embedded by a watermark in the fingerprinting applications are used to trace the originator or recipients of a particular copy of a multimedia file. The usage of an audio file can be recorded by a fingerprinting system. When a file is accessed by a user, a watermark, or called fingerprint in this case, is embedded into the file thus creating a mark on the audio. The usage history can be traced by extracting all the watermarks that were embedded into the file

[VI]Broadcast monitoring A variety of applications for audio watermarking are in the field of broadcasting. Watermarking is an obvious alternative method of coding identification information for an active broadcast monitoring. It has the advantage of being embedded within the multimedia host signal itself rather than exploiting a particular segment of the broadcast signal. Thus, it is compatible with the already installed base of broadcast equipment, including digital and analogue communication channels. [VII]Medical applications Watermarking can be used to write the unique name of the patient on the X-ray reports or MRI scan reports. This application is important because it is highly advisable to have the patients name entered on

reports, and reduces the misplacements of reports which are very important during treatment. [VIII] Airline traffic monitoring Watermarking is used in air traffic monitoring. The pilot communicates with a ground monitoring system through voice at a particular frequency. However, it can be easily trapped and attacked, and is one of the causes of miss communication. To avoid such problems, the flight number is embedded into the voice communication between the ground operator and the flight pilot. As the flight numbers are unique the tracking of flights will become more secure and easy.

1.5 Requirement of watermarking According to IFPI (International Federation of the Phonographic Industry) audio watermarking algorithms should meet certain requirements. The most significant requirements are perceptibility, robustness, security, reliability, capacity, and speed performance.

[1] Imperceptibility: One of the important features of the watermarking technique is that the watermarked signal should not lose the quality of the original signal. The signal to noise ratio (SNR) of the watermarked signal to the original signal should be maintained greater than 20dB. In

addition, the technique should make the modified signal not perceivable by human ear.

[2]Robustness to signal processing attacks: Watermark should be robust against common signal processing attacks such as lousy compression, linear filtering, Re sampling, Re quantization, cropping, jittering, D/A and A/D conversion, Normal correlation (N.C.) is used for the measurement of robustness.

[3] Capacity: The efficient watermarking technique should be able to carry more information but should not degrade the quality of the audio signal. It is also important to know if the watermark is completely distributed over the host signal because, it is possible that near the extraction process a part of the signal is only available. Hence, capacity is also a primary concern in the real time situations. [4]Speed: Speed of embedding is one of the criteria for efficient watermarking technique. The speed of embedding of watermark is important in real time applications where the embedding is done on continuous signals such as speech of an official or conversation between airplane pilot and ground control staff. Some of the possible applications where speed is a constraint are audio streaming and airline traffic monitoring. Both embedding and extraction process need to be made as fast as possible with greater efficiency.

[5] Security and cost: Cost is an important criterion which cannot be ignored and watermark should be secure so that no one can know about our watermark.The security of a watermark refers to its ability to resist hostile attacks. Hostile attack is the process specifically intended to thwart the watermark’s purpose. The types of attacks can fall in three categories: unauthorized removal, unauthorized embedding, and unauthorized detection. The Cost of watermarking system refers to the speed with which embedding and detection must be performed and the number of embedded and detectors that must be deployed. Other issues include the whether the detector and embedded are to be implemented as hardware device or as software application or plug-ins.

[6]Computational complexity: Computational complexity refers to the processing required to embed watermark data into a host signal, and /or to extract the data from the signal. Algorithm complexity is important to know, for it may influence the choice of implementation structure or DSP architecture. Although there are many ways to measure complexity, such as complexity analysis (or “Big-O” analysis), for practical applications more quantitative values are required . In this study, actual CPU timings (in seconds) of algorithm implementations were collected.

[7] Asymmetry: If for the entire set of cover objects the watermark remains same then extracting for one file will cause damage watermark of all the files. Thus, asymmetry is also a noticeable concern. It is recommended to have unique watermarks to different files to help make the technique more useful.

1.6 Problems and attacks on watermarking[1] Compression: compression means to reduce the size of signal. Compression of audio signal is a very common signal processing tool. Audio files are compressed when these files are stored at disk at server or transmitted over a communication channel. Watermark should be robust against compression. Audio generation is done at a particular sampling frequency and bit rate however the created audio track will undergo so many different types of compression and conversion techniques. Some of the most common compression techniques are audio compression techniques based on psychoacoustic effect (MPEG and Advanced Audio Codec (AAC)). In addition to that, it is common process that the original audio signal will change its sampling frequencies like from 128Kbps to 64Kpbs or 48 Kbps. There are some programs that can achieve these conversions and perform compression operation.

[2] Re sampling: During Analog to digital conversion a signal can be Re sampled at different -different sampling frequencies. when an audio signals converted from analog to digital first it sampled at Nyquest rate these samples are quantized and then encoded into digital signals and again when these digital signals are converted back to analog signal first quantized than decoded into an analog signals and then low pass filtered to produce an original audio signals During various signal processing Techniques a signal may be re sampled at different rates(or when signal go through an intentional attack) so the watermark should be embedded such that it will not effected more by re sampling attack.

[3] Re quantization: A signal can be Re quantized with different quantization Level as we change the quantization level the quantization error also changed and no of bits used to represent per sample also changed.During analog to digital conversion of an audio signal an audio signal is first sampled at Nyquest rate than quantized during the process of quantization a quantization error is introduced quantization error is simply the difference between the sampled value of the signal and quantized value of the signal. The Quantization error is inversely proportional to the no of quantization level if we increase the no of quantization level then quantization error will be decreased and if we decrease the no of quantization level than

quantization error increase. Again as we increase the no of quantization level we do require a more no of bits to represent the quantized sample and as we decrease the no of quantization level we do require a less no of bit to represent the quantized value of sample. Hence there is a tradeoff between quantization error and no of bits to encode the sample i.e. if we increase the no of bits then quantization error decrease and if we decrease the no bits than quantization error increase. A signal may go through a quantization attack during various signal processing and watermark should be embedded such that it remain unaffected (less effected) after re quantization attack.

[4] Filtering: Filtering is used to filter certain frequency components of any signal or we can say that it attenuates certain frequency component of any signal. Filtering is common practice, which is used to amplify or attenuate some part of the signal. The basic low pass and high pass filters can be used to achieve these types of attacks.

[5] Additive white Gaussian noise attacks: A.W.G.N. can be added to watermark signal during attack. It is common practice to notice the presence of noise in a signal when transmitted. Hence, watermarking algorithm should make the technique robust against the noise attacks. It is recommended to check the algorithm for this type of noise by adding the host signal by an additive

white Gaussian noise (AWGN) to check its robustness.

[6]Cropping: Cropping means to crop the some portion of the signal or replace with an another signal the watermark should be distributed over the signal such that it will effected very less by cropping attack.

[7] Dynamics: The amplitude modification and attenuation provide the dynamics of the attacks. Limiting, expansion and compressions are some sort of more complicated applications which are the non-linear modifications. Some of these types of attacks are re-quantization.

[8] Ambience: In some situations the audio signal gets delayed or there are situations where in people record signal from a source and claim that the track is theirs. Those situations can be simulated in a room, which is of great importance to check the performance of an audio signal.

[9] Time stretch and pitch shift: These attacks change either the length of the signal without changing its pitch and vice versa. These are some de-synchronization attacks which are quite common in the data transmission. Jittering is one type of such attack. Embedded in the host signal itself which is one of the main advantages of the technique.

1.7 Performance Parameters There are following performance parameters which we are using[1] Signal to noise ratio for Perceptual Quality[2] Normalized Correlation for Robustness1.7.1 Signal to noise ratio (S.N.R.): The S.N.R. is used for the measurement of perceptual quality (inaudibility) of watermarked signal. In order to evaluate the imperceptibility of watermarked signal, the following signal-to-noise ratio (SNR) equation is used.

……………………………… (1.1)Where S (n) is a host audio signal of length N samples and S′ (n) be watermarked audio signal.Subjective Quality evaluationThe inaudibility of our watermarking method has been done by listening tests involving ten persons. Each listener was presented with the pairs of original signal and the watermarked signal and was asked to report whether any difference could be detected between the two signals. The ten people

listed to each pair for 10 times and they have a grade for this pair, using the ITU-R BS.1284 standardized 5-point grading scale [15]. The average grade for of each pair from all listeners is the final grade for this pair. Table 1. Grading Scale (ITU-R)

Grade Quality Impairment5 Excellent Imperceptible4 Good Perceptible but

not annoying3 Fair Slightly annoying2 Poor Annoying1 Bad Very Annoying

1.7.2 Normalized correlationNormalized correlation (N.C.) is used for the measurement of robustness of watermark against various signal processing attacks like Re sampling, Re quantization, Compression, low pass filtering, cropping etc. N.C. (normal correlation) can be a more sensible measure for expressing the robustness of the audio watermarking algorithm. Correlation is the measure of similarity between two signals. Correlation can be measured by using normalized signal which is termed as normalized correlation. Normal correlation can be calculated as

……………………. (1.2)

In above equation w and w’ are original and extracted watermarks and i and j are indexes of watermark image. In fact, by setting a threshold value for NC, the receiver can decide whether the extracted watermark correlates (is similar) with the signature embedded watermark.

CHAPTER-2 LITRATURE REVIEW

2.1 History of watermarking:The term "digital watermark" was first coined in 1992 by Andrew Tirkel and Charles Osborne. Actually, the term used by Tirkel and Osborne was originally used in Japan-- from the Japanese-- "denshi sukashi" -- literally, an "electronic watermark". Before the invention of steganography and cryptography, it was challenging to transfer secure information and thus to achieve secure communication environment. Some of the techniques employed in early days are writing with an invisible ink, drawing a standard painting with some small modifications, combining two images to create a new image, shaving the head of the messenger in the form of a message, tattooing the message on the scalp and so on. 2.2 Summary of literature reviewWhile carrying out the work related to this thesis work. I have reviewed the various Paper the various papers are summarized below in table form

S.N.

NAME OF AUTHOR

RESEARCH PAPER NAME

NAME OF JOURNALS / YEAR OF PUBLICATION

RESEARCH AREA

1. [1]Himeur Yassine

A SECURE AND HIGH ROBUST

International Journal of

DWT+DSTwatermark is used

Department of electronics University of Algeria ,[2]Khelalef Aziz Department of electronics University of Batna Algeria

AUDIO WATERMARKING SYSTEM FOR COPYRIGHT PROTECTION

Computer Applications (0975 – 8887) Volume 53– No.17, September 2012

for modifying the D.S.T.(discrete sine coefficient) coefficient of an audio signal

2. [1]Mohammad Ibrahim [2]Kaushik Deb

A NEW AUDIO WATERMARKING METHOD BASED ON DISCRETE CO- SINE TRANSFORM WITH A GRAY IMAGE

International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 4,August 2012

D.C.T.Watermark signal (gray image) is added at Peak prominent value of D.C.T. coefficient of an audio signal

3 [1] Ali Al-aj[2]Ahmad Mohammad[3] Lama Bata

DWT BASED AUDIO WATERMARKING

The International Arab Journal of Information Technology, Vol. 8, No. 3, July 2011

Watermark is added in Discrete Wavelet Domain of an audio signal at most appropriate frequency bands

4 [1] Xu Yan-ping [2] Jia Li-qin

RESEARCH OF A DIGITAL WATERMARKING ALGORITHM BASED ON DISCRETE COSINE TRANSFORM

Proceedings of the Third International Symposium on Electronic Commerce and Security Workshops(ISECS ’10) Guangzhou P R China 29-31 July

DCTA random sequence is used as a watermark signal embedded with the DCT coefficient of an audio signal

2010 pp 373-375

5 [1]Yasunari Yoshitomi [2]Yohei Kinugawa Kyoto Prefectural UniversityKyoto Japan

AN AUTHENTICATION METHOD FOR DIGITAL AUDIO USING A DISCRETE WAVELET TRANSFORM

Journal of Information Security, April 2011

DWT Information is added with DWT coefficient of an audio signal for authentication proposes. and added watermark is robust against various signal processing attacks

6 [1] M. O Agbaje [2] A.T Akinwale [3]A.N Njah

AUDIO WATERMARKING: A CRITICAL REVIEW

International Journal of Scientific & Engineering Research, Volume2,I

This paper compare different audio watermarking algorithm and shows advantage

ssue 11, November-2011

s and disadvantages of one algorithm over other and also shows the suitability of different methods for different-different applications.

7 [1] De Li , [2]Yingying Ji [3]JongWeon Kim

AUDIO WATERMARKING BY COEFFICIENT QUANTIZATION IN THE DWT-DCT DUAL DOMAIN

International Journal of Security and Its Applications Vol.7, No.5 (2013),

The DCT coefficient of low frequency DWT coefficients are quantized with the help of an

pp.183-192

image signal.

8 [1] Hooman Nikmehr, [2] Sina Tayefeh Hashemy

A NEW APPROACH TO AUDIO WATERMARKING USING DISCRETE WAVELET AND COSINE TRANSFORMS

International conference on communication engineering 22-24 dec 2010 University of sistan & Salicheston

DWT+DCTAudio signal is segmented into two equal size groupThe synchronization information is hide in dwt coefficient of first segmentThe watermark information is hide in DWT+DCT coefficient of second segment

The technique which are proposed above are not much robust against signal processing Attacks such as compression and low pass filtering particularly for the low value of α (scaling factor decide the strength of watermarking) . Again if we increase the values of α for improving the robustness then the perceptual Quality of the signal degraded. Again some techniques described above are much complex and are not cost effective.

2.3 Problem Formulation Robustness and Imperceptibility are main Requirement of an Audio watermarking again there is a tradeoff between above Requirements. if we try to improve one performance parameter and we need to do Compromise with other performance parameter. On reviewing the literature it is clear that the existing algorithm for audio watermarking are not much robust against various intentional and unintentional signal processing attacks particularly for the low value of α. Again if we increase the values of α for improving the robustness then the perceptual Quality of the signal degraded.For our proposed algorithm the perceptual quality of watermarked signal is very high. Again the proposed algorithm is very robust against various intentional or unintentional common signal processing attacks such as Compression, Low pass filtering, Re sampling, Re quantization, Cropping, Adding Gaussian white noise etc.

CHAPTER-3 THEORY OF WATERMARKING

3.1 Introduction “Watermarking is the process of embedding information into a signal (for example audio, video or pictures) in a way that is difficult to remove. If the signal is copied, then the information is also carried in the copy.” 3.2 Different algorithm for audio watermarking Almost all audio watermarking algorithms take advantage of the weakness of HAS or the

perceptual properties of HAS to embed the watermark data in the parts of audio signals so that the distortion resulted from the watermarking process is not audible. The most common algorithms are

1. L.S.B. Coding 2. Echo Watermarking 3. Patchwork Techniques 4. Quantization index modulations 5. Phase coding 6. Spread spectrum modulations

Further the watermarking can be done in time domain as well as in frequency domain.The commonly used time domain algorithms are LSB Coding, Echo hiding, patchwork algorithms etc.Whereas frequency domain algorithms are Phase coding, Coefficient quantization embedding the watermark by modifying the coefficients of the transform outputs such as F.F.T. (Fast Fourier Transform)S.T.F. T. (Sort time Fourier Transform) D. C. T. (Discrete Fourier Transform)D.W. T. (Discrete Wavelet Transform) etc

3.2.1 L.S.B. Coding This technique is one of the common techniques employed in signal processing applications. It is based on the substitution of the LSB of the carrier signal with the bit pattern from the watermark

noise. The robustness depends on the number of bits that are being replaced in the host signal. This type of technique is commonly used in image watermarking because each pixel is represented as an integer hence it will be easy to replace the bits. The audio signal has real values as samples if converted to an integer will degrade the quality of the signal to a great extent. The operation of the 2-bit LSB coding is shown in Figure.

Fig 3.1 L.S.B. Coding

3.2.2 Echo watermarkingEcho hiding schemes embed watermarks into a host signal by adding echoes to produce watermarked signal.The nature of the echo is to add resonance to the host audio. After the echo has been added,

Fig 3.2 Echo watermarkingWatermarked signal retains the same statistical and perceptual characteristics. The offset (or delay) between the original and a watermarked signal is small enough that the echo is perceived by the HAS as an added resonance.

3.2.3 Patchwork Technique The data to be watermarked is divided into two distinct subsets. One feature of the data is selected and customized in opposite directions in both subsets . For an example let the original signal is divided into two parts A and B, then the part A is increased by a fraction Δ and the part B is decreased by some amount Δ. The samples separation is the secret key which is termed as watermarking key.Let NA and NB indicate the size(s) of the individual A and B parts and Δ be the total of the change made to the host signal. Suppose that a[i] and b[i] represent the sample values at i position in blocks A and B. The difference of the sample values can be written as

…………. (3.1)The expectation of the difference is used to extract the watermark which is expressed as follows.

…..…...... (3.2)3.2.4 Quantization index modulation The quantization index modulation (QIM) is a technique which uses quantization of samples to embed watermark.

Fig 3.3 Modification of sample using Q.I.MThe basic principle of QIM is to find the maximum value of the samples and to divide the range 0 to the maximum value into intervals of step size Δ. The intervals are assigned a value of 0 or 1 depending on any pseudo random sequence. Each sample has quantized value, thus, a polarity is assigned based on the location of the interval. The watermark is embedded by changing the value of the median for created interval and by the similarity of the polarity and watermark bit.

3.2.5 Phase coding [Watermarking the phase of the host signal] Algorithms that embed watermark into the phase of the host audio signal do not use masking properties of the HAS, but the fact that the HAS is insensitive to a constant relative phase shift in a stationary audio signal. There are two main approaches used in the watermarking of the host signal’s phase.1. Phase coding2. Phase modulation1. Phase coding: In phase coding method. The basic idea is to split the original audio stream into blocks and embed the whole watermark data sequence into the phase spectrum of the first block. One drawback of the phase coding method is a considerably low payload because only the first block is used for watermark embedding. In addition, the watermark is not dispersed over the entire data set available, but is implicitly localized and can thus be removed easily by the cropping attack. It is a non-blind watermarking method (as the phase modulation algorithm) that limits the number of applications it is suitable for.2. Phase modulation: The watermark insertion in the phase modulation method is performed using an independent multiband phase modulation. Imperceptible phase modifications are exploited in this approach by the controlled phase alternation of the host audio. To ensure perceptual transparency by introducing only small changes in

the envelope, the performed phase modulation has to satisfy the following constraint

o …………………… (3.3)where Ө(z) denotes the signal phase and z is the Bark scale. Each Bark constitutes one critical bandwidth. Using a long block size N (e.g. N = 214) algorithm attains a slow phase change over time. The watermark is converted into a phase modulation by having one integer Bark scale carry one message bit of the watermark, with the frequency in Hz. The robustness of the modulated phase can be increased by using multiple Bark values carrying one watermark bit.The watermark extraction requires a perfect synchronization procedure to perform a block alignment for each watermarked block, using the original signal as a reference. A matching of the particular segments of the modulated phase to the encoded watermark bits is possible if no significant distortions of the watermarked signal took place. The data rate of the watermark depends on three factors: first, the amount of the redundancy added, second, the frequency range used for watermark embedding, and, third, the energy distribution of the host audio. 3.2.6 Spread spectrum watermarking algorithmSpread-spectrum watermarking scheme is an example of the correlation method which embeds pseudorandom sequence and detects watermark

by calculating correlation between pseudo-random noise sequence and watermarked audio signal.Basic ideaThe modulated signal is then added to the original audio to produce the watermarked audio x(n) such as X (n) = s (n) + α .w (n) …..……………….. (3.4) The detection scheme uses linear correlation. Because the pseudo-random sequence r (n) is known and can be regenerated generated by means of a secret key, watermarks are detected by using correlation between x (n) and r (n) such as ………………............ (3.5) Where, N denotes the length of signal. Equation yields the correlation sum of two components as follows:

Fig 3.4 Detection process of spread spectrum watermarking

3.3 Domain of watermarkingThe watermark can be embedded with host audio signal either in time domain (temporal domain) or in frequency domain (spectral domain) .So there are two domain of audio watermarking[I] Time domain watermarking (Temporal watermarking)[2] Frequency domain watermarking (Spectral watermarking)

3.3.1 Time domain watermarking (Temporal watermarking).Temporal watermarking hides watermarks directly into digital audio signals in the time domain, time-domain audio watermarking is relatively easy to implement, and requires few computing resources, low cost, fast speed however, it is weak against signal processing attacks such as compression and filtering.In time domain watermarking techniques, watermark is directly embedded into audio signal. No domain transform is required in this process. Watermark signal is shaped before embedding operation to ensure its inaudibility.

Fig 3.5 Time domain audio watermarking

Watermark ShapingCarelessly added pseudo-random sequence or noise to audio signal can cause unpleasant audible sound whatever watermarking schemes are used. Thus, just reducing the strength α of pseudo-random sequence cannot be the final solution. Because human ears are very sensitive especially when the sound energy is very low, even a very little noise with small value of α can be heard. Moreover, small α makes the spread-spectrum scheme not robust. One solution to ensure inaudibility is watermark shaping based on the psycho-acoustic model. Interestingly enough, the watermark shaping can also enhance robustness since we can increase the strength α sufficiently as

long as the noise is below the margin. Psycho-acoustic models for audio compression exploit frequency and temporal masking effects to ensure inaudibility by shaping the quantized noise according to the masking threshold. However, masking effect can increase the minimum masking threshold. A sound lying in the frequency or temporal neighborhood of another sound affects the characteristics of the neighboring sound, which phenomenon is known as masking. The sound that does the masking is called masker and the sound that is masked is called the masque. The psycho-acoustic model analyzes the input signal s(n) in order to calculate the minimum masking threshold T.

Fig 3.6 Watermark shaping

3.3.2 Frequency domain audio watermarking

The input signal is first transformed to frequency domain where the watermark is embedded, the resulting signal then goes through inverse frequency transform to get the watermarked signal as output as shown in Figure .In the transform domain. FFT, DCT, DWT, are commonly used transform. Frequency domain watermarking is very robust again various signal processing attacks. Audio watermarking techniques that work in frequency domain take the advantage of audio masking characteristics of HAS to embed an inaudible watermark signal in digital audio. Transforming audio signal from time domain to frequency domain enables watermarking system to embed the watermark into perceptually significant components. This will provide the system with a high level of robustness [Cox et al, 1997] because of that any attempt to remove the watermark will result in introducing a serious distortion in original audio signal fidelity. The input signal is first transformed to frequency domain where the watermark is embedded, the resulting signal then goes through inverse frequency transform to get the watermarked signal as output as shown in Figure .

Fig 3.7 Frequency domain audio watermarking In order for the watermark to be robust, watermark must be placed in perceptually significant regions of the cover signal despite the risk of potential fidelity distortion. Conversely if the watermark is placed in perceptually insignificant regions it is easily removed either intentionally or unintentionally for example, signals compression techniques that implicitly recognize that perceptually weak components of a signal need not be represented. 3.4 The human Audi oratory system [H.A.S.] Watermarking of audio signals is more challenging compared to the watermarking of images or video , due to wider dynamic range of the HAS in comparison with human visual system (HVS) .The HAS perceives sounds over a range of power greater than 109:1 and a range of frequencies greater than 103:1. The sensitivity of the HAS to the additive white Gaussian noise (AWGN) is high as well noise in a sound file can be detected as low as 70 dB below ambient level. Most audio

watermarking schemes rely on the imperfections of the human auditory system (HAS). HAS is insensitive to small amplitude changes in the time and frequency domains, allowing the addition of weak noise signals (watermarks) to the host audio signal such that the changes are inaudible. In the time domain, it has been demonstrated that the HAS is insensitive to small level changes and insertion of low-amplitude echoes [6]. Data hiding in the frequency domain takes advantage of the insensitivity of the HAS to small spectral magnitude changes [6].Further, HAS is insensitive to a constant relative phase shift in a stationary audio signal and some spectral distortions are interpreted as natural, perceptually non-annoying ones [7].Two Properties of the HAS dominantly used in watermarking algorithms are1. Frequency (simultaneous) masking 2. Temporal masking3.4.1 Frequency (simultaneous) masking: Frequency (simultaneous) masking is a frequency domain observable fact where low levels signal (the masque) can be made inaudible (masked) by a simultaneously appearing stronger signal (the masker), if the masker and masque are close enough to each other in frequency [8]. A masking threshold can be found and is the level below which the audio signal is not audible. A masking threshold can be derived below which any

signal will not be audible. The masking threshold depends on the masker and on the characteristics of the masker and masque (narrowband noise or pure tone). For example, with the masking threshold for the sound pressure level (SPL) equal to 60 dB, the masker at around 1 kHz. The SPL of the masque can be surprisingly high it will be masked as long as its SPL is below the masking threshold. The slope of the masking threshold is steeper toward lower frequencies; in other words, higher frequencies tend to be more easily masked than lower frequencies. As shown in figure.It should be pointed out that the distance between masking level and masking threshold is smaller in noise-masks tone experiments than in tone-masks-noise experiments due to HAS’s sensitivity toward additive noise. Noise and low-level signal components are masked inside and outside the particular critical band if their SPL is below the masking threshold. Noise contributions can be coding noise, inserted watermark sequence, aliasing distortions, etc.

Figure 3.8 Frequency masking in the human auditory system (HAS)

The distance between the level of the masker (given as a tone in Fig) and the masking threshold is called signal-to-mask ratio (SMR). Its maximum value is at the left border of the critical band. Within a critical band, noise caused by watermark embedding will be audible as long as signal-to-noise ratio (SNR) for the critical band is higher than its SMR. Let SNR (m) be the signal-to-noise ratio resulting from watermark insertion in the critical band m; the perceivable distortion in a given sub band is then measured by the noise to mask ratio: NMR (m) = SMR-SNR (m) ……….……………… 3.6The noise-to-mask ratio NMR (m) expresses the difference between the watermark noise in a given critical band and the level where a distortion may just become audible; its value in dB should be negative. This description is the case of masking

by only one masker. If the source signal consists of many simultaneous maskers, a global masking threshold can be computed that describes the threshold of just noticeable distortion (JND) as a function of frequency.3.4.2 Temporal masking:.In the time domain, it has been demonstrated that the HAS is insensitive to small level changes and insertion of low-amplitude echoes [9]. In addition to frequency masking, two phenomena of the HAS in the time domain also play an important role in human auditory perception. Those are pre-masking and post-masking in time. The temporal masking effects appear before and after a masking

Fig 3.9 Temporal Masking of Human Auditory systemSignal has been switched on and off, respectively (Fig). The duration of the pre- masking is significantly less than one-tenth that of the post-

masking, which is in the interval of 50 to 200 milliseconds. Both pre- and post-masking have been exploited in the MPEG audio compression algorithm and several audio watermarking methods.Frequency domain techniques, in particular, have been more effective than time-domain techniques since watermarks are added to selected regions in the transformed domain of the host audio signal, such that inaudibility an robustness are maintained [10]. CHAPTER-4 TRANSFORMATION TECHNIQUES Any signal in time domain is considered as raw signal. The Propose of all Transformation techniques are to convert time domain signal in a form so that desired information can be extracted from these signal’s and after the application of certain transform the resultant signal is known as processed signal.The commonly used transformation techniques are Discreet Fourier Transform Short time Fourier transform Wavelet transform4.1. Discrete Fourier transform: The Discrete Fourier transform is used for converting time domain signal into frequency domain In time domain representation of the signal there is a graph between time and amplitude. The time domain representation of the signal gives information about the signal that at which time instants what is the amplitude of the signal or how amplitude of the

signal is varying with respect to time but it gives no information about the Different frequency contents that are presents in the signal. In frequency domain representation of the signal there is a graph between frequency and amplitude .The frequency domain representation of the signal tells us that in a given signal what different frequency components are present and what respective amplitudes are there. But again frequency domain representation gives no idea that at which time these frequency components are presents. In some applications frequency domain representation is more important. Or we can say that frequency domain representation gives more information about any signal (for example audio music signal). Suppose x= 100*sin (2*pi*2*t) +50*cos (2*pi*3*t) is a time Domain representation of a given signal andy= fft (x) is frequency domain representation of signal x

0 1 2 3 4 5 6 7 8 9 10-200

-100

0

100

200

TIME

AMPL

ITUD

E

0 1 2 3 4 5 6 7 8 9 10-1000

0

1000

2000

3000

FREQUENCY

AMPL

ITUD

ETIME DOMAIN REPRESENTATION OF SIGNAL

FREQUENCY DOMAIN REPRESENTATION OF SIGNAL

Fig 4.1 Time Domain and frequency Domain representation of signalAs it is clear from fig 4.1 that time domain representation of signal gives no idea about the frequency components of signal it simply shows that how the amplitude of the signal is varying with respect to time whereas frequency domain representation of the signal shows the different frequency components presents in a signal and their respective amplitudes but it gives no idea that at which time instants these frequency components presents.In the other word we can say in time domain representation of signal the time resolution of signal is very high but its frequency resolution is zero because it gives no idea about different frequency components presents in a signal where

as in frequency domain representation of the signal the frequency resolution of the signal is very high but its time resolution is zero because it gives no idea about time.Both domain of signal analysis has its own utilities and has its own importance and having its own sets of advantages and disadvantages. How Fourier transform convert time domain signal into frequency domain. Suppose x (t) shows the time domain representation of signal and X (f) shows the frequency domain signal Then

…………………………………………….. (4.1)

……………………………………………. (4.2)Where

) + j Sin ) …………………………………. (4.3)

Equation (4.1) and equation (4.2) gives Fourier transform and inverse Fourier transform of any signal the exponential term of equation (1) can be expressed in terms of sine and cosine function shown by equation (4.3)With the help of equation (4.1) and equation (4.2) we can convert any time domain signal into frequency domain signal and frequency domain signal into time domain signal respectively.As clear from equation (4.1) that in Fourier transform the signal is integrated from – infinity to

+ infinity over time for each frequency In the other word we can say that equation 1 take a frequency for example f1 and search it from –infinity to + infinity over time if it find the f1 frequency components it simply adds the magnitude of all f1 frequency components.Again take an another frequency for example f2 and search it from – infinity to + infinity over time if it find the f2 frequency components it simply adds the magnitude of all f2 frequency components Again repeat the same process with f3,f 4, f 5……. and so on No matter in time axis where these frequency components exits from – infinity to + infinity it will affect the result of integration in the same way For every frequency Fourier transform check that whether this particular frequency component present or not present in time from minus infinite to plus infinite. And if present then how many time these particular frequency components presents and what is the amplitude of this particular frequency component and then simply add that particular frequency component and calculate the amplitude of any particular frequency component. Again take a second frequency component and check that in time from minus infinite to plus infinite how many times this particular frequency component exists and what is amplitude of this particular frequency component and then simply adds them. In this way the Fourier transform calculated the amplitude of every frequency

components presents in a given signal and draw a graph between frequency and amplitudeAgain there are certain disadvantage of frequency domain representation of the signal first disadvantage is that it gives no idea about time .The frequency transform of any signal simply tells us that in any given signal what spectral components are present and what are their respective amplitudes but it gives no idea that in time axis where these frequency components existsSo again the D.F.T. prove its suitability for the signals which are stationary in nature but this transform is not suitable for non stationary signalBy stationary signal we simply means the signal in which the frequency does not change with respect to time or we can say that all frequency components exits for all the time By non stationary signal we simply mean the signal in which frequency changes with respect to the time. or in which all the frequency components does not exist for all the time interval .but some frequencies are exits for some particular time interval whereas some other frequency components exists for some other time interval.To understand the suitability of the DFT only for stationary signal and not for non stationary signal take the following example suppose there are two signals S1 and S2 the signal S1 has three frequency components f1,f2 and f3 all the times and suppose the signal S2 contains the same frequency components f1,f2 and f3 but for the different -

different time interval’s so we can say that these two signals are completely different in nature. But in spite of it the Fourier transform of these two signals will be the same because these two signals have the same frequency components of course one signal contains all the frequency components all the time whereas second signal contains these frequency components at different time intervals but as we know that Fourier transform has nothing to do with the time. No matter where these frequency components exits over time the matter is only that whether they occur or not and what are their amplitudes.

Fig 4.2 Time domain and frequency domain representation of stationary and non stationary signalsAgain take an another example suppose there are two signals S3 and S4.Signal S3 contains frequencies f1 for time interval t0 to t1, frequency f2 for time interval t1 to t2, and frequency f3 for time interval t2 to t3.And signal S4 contain frequency f3 for time interval t0 to t1, frequency f2 for t1 to t2 and frequency f1 for time interval t2 to t3 So we can say that these two signals are quiet different though both signals are having the same frequency components in same amount but the time instances where these frequency components exists are different so the overall characteristics of above these two signals S3 and S4 will be different but in spite of this the Fourier transform of these two signals will be the same because Fourier transform has nothing to do with time.

Fourier transform simply watch that what frequency components any signal has and what are their respective amplitudes

4.2 Short time Fourier transformsThe Short time Fourier Transform is a modified version of Fourier transform. S.T.F.T. is nothing it is simply the Fourier transform of any signal multiplied by a window function.

STFTX (w) (t, f) = ∫t [x(t). w*( t – t')].e-j2Πf t dt ………………………..4.4

The basic idea behind the STFT is that any non stationary signal can be considered stationary for a short time interval. So we can say that the STFT gives an idea about time frequency and amplitude.But again the problem with STFT is that how to choose the size of window (time interval of window) because .Any single window size is not suitable for the analysis of all frequency components small window size is suitable for the analysis of high frequency components whereas the large window size is suitable for the small frequency component.

Fig (4.3) STFT has same window size for all frequency component But the problem with STFT is that the window size remain same for the all analysis we can’t choose different -different window Size for the analysis of different frequency components .As shown in fig that for all frequencies the size of window is same Hence we can say that there is some resolution problem with S.T.F.T. at the same time we can’t obtain both time as well as frequency resolution means we cannot know Exactly at which particular time instant which particular frequency components Exists We can know only that in which time intervals(not time instants) what frequency spectrum occurs(not exact frequency)4.3 Wavelet transform

…………………… (4.5)By using wavelet Transform we can overcome the problem with S.T.F.T. in wavelet transform we used different window size for different frequency components. Low scale (small window size or small time scale) is used for high frequencies and high scale (Large Window size or large time scale) is used for low frequencies 4.3.1 What is Multi resolution property?Multi resolution property means different frequency components presents in any signal are resolved at different scale(different time scale or different window size).Scale is inversely proportional to the frequency means small scale is used for higher frequencies whereas large scale is used for small frequenciesSmall scale (Small window Size or Small time scale)is used for the analysis of higher frequency components. Means wavelet transform provide higher time resolution for high frequency means if any signal contain a very high frequency component then with the help of wavelet transform we can know that at which exact time interval (very small time interval) these frequency components exists.whereas the large scale (large window size or large time scale)is used for the analysis of small frequency components .Wavelet transform provide good time resolution for higher frequencies

whereas for small frequencies time resolution is not so good .But if we study real word signals then we find that generally higher frequencies are occurs only for very short time interval whereas small frequency components are presents for long time interval. So wavelet transform prove its suitability for real word signals.4.3.2 What is spectral localization?Spectral localization property means that wavelet transform tells us that what frequency components are present in any given signal and at time axis where these frequency components are presents

4.3.3 What is Scale?Scale is inversely proportional to the frequency. Large scale is used for the analysis of small frequency components presents in any signal Whereas Small scale is used for the analysis of high frequency components of any signal

Fig (4.4) DWT Uses different scale for different frequency components

4.4 Process of wavelet transformsIn the process of wavelet transform the original signal(S) is first decompose into Approximate Coefficients and Detailed Coefficients by simply passing the signal through low pass filter and high pass filter respectively.The output of low pass filter is called Approximate [A1] (Low frequency components) coefficient of the signalThe output of High pass filter is called Detailed [D1] (High frequency components) Coefficients Of the signal.This Approximate coefficient [A1] again passed through a low pass and high passes filter and again Decompose the signal into Approximate [A 2] and Detailed Coefficients [D 2]Further Approximate Components [A2] can be decomposed into approximate coefficients [A3] And Detailed Coefficients [D3]

The number of Decomposition levels depends on the length of signal and our requirementsS = A1+D1 [First level Wavelet Decomposition]…………………………….. (4.6)

A1 = A2+D2 [Second level Wavelet Decomposition]……................................... (4.7)

A2 = A3+D3 [Third level Wavelet Decomposition]………................................. (4.8)

S = A3+D3+D2+D1 ……………………………………………………………….. (4.9)

The Original Signal S can be reconstructing with the help of A3, D3, D2 and D1.

Fig (4.5) Wavelet Decomposition of signal S=A3+D3+D2+D1

Signal

(s)

A-1D-1

D-2 A-2

A-3D-3

So it is clear that with the help of equation (4.6) and equation (4.7), (4.8), (4.9) We can decompose any original signal sequences in to wavelet decomposition. And with these wavelet decomposition again we can construct the original signal. The number of sample in next decomposition level is half as compared to previous stage. Suppose The original signal S has N samples then A1 and D1 will have N/2 Samples and A2 and D2 will have N/4 Samples.

4.5 THE DISCRETE COSINE TRANSFORM DCT is a transform representing a signal in the form of a series of coefficients obtained from a sum of cosine functions oscillating at different frequencies and at different amplitudes [12].DCTs are important to numerous applications in science and engineering, from lossy compression of audio (e.g. MP3) and images (e.g. JPEG) (where small high-frequency components can be discarded), to spectral methods for the numerical solution of partial differential equations. The use of cosine rather than sine functions is critical in these applications: for compression, it turns out that cosine functions are much more efficient (as described below, fewer coefficients are needed to approximate a typical signal), whereas for differential equations the cosines express a particular choice of boundary conditions [13].

DCT is a Fourier-related transform similar to the discrete Fourier transform (DFT), but using only real numbers. DCTs are equivalent to DFTs of roughly twice the length, operating on real data with even symmetry (since the Fourier transform of a real and even function is real and even), where in some variants the input and/or output data are shifted by half a sample. There are eight standard DCT variants, of which four are common. The most common variant of discrete cosine transform is the type-II DCT, which is often called simply "the DCT"; its inverse, the type-III DCT, is correspondingly often called simply "the inverse DCT” or "the IDCT".The importance of DCT II is further accentuated by its -

Superiority in bandwidth compression (redundancy reduction) of a wide range of signals.

Powerful performance in the bit-rate reduction. Existence of fast algorithms for its

implementation.DCT-II and its inversion, DCT-III, have been employed in the international image/video coding standards: e.g.: JPEG, MPEG, H.261, H.263, H.264… Like other transforms, the Discrete Cosine Transform (DCT) attempts to décor relate the image data. After décor relation each transform coefficient can be encoded independently without losing compression efficiency. The most common DCT definition of a 1-D sequence of length N is [13].

………………………. 4.10

For k= 0, 1, 2 … N-1. Similarly, the inverse transformation is defined as:

………………………. 4.11

For n= 0, 1, 2 …N-1. In both equations (4.10) and (4.11) w (k) is defined as:

…………………………4.12

Ability in compressing energy of the signal in few coefficients is one of the criteria for comparing performance of the transforms. DCT is among the best in term of the compressing capability and therefore, when quantizing, the transform is allowed to ignore the coefficients with low amplitudes without losing the accuracy during reconstructing the signal from its coefficients.The first transform coefficient is the average value of the sample sequence. In literature, this value is referred to as the DC Coefficient. All other transform coefficients are called the AC Coefficients.Compared with DFT, DCT has two main advantages:

It’s a real transform with better computational efficiency than DFT which by definition is a complex transform. It does not introduce discontinuity while imposing periodicity in the time signal. In DFT, as the time signal is truncated and assumed periodic, discontinuity is introduced in time domain and some corresponding artifacts are introduced in frequency domain. But as even symmetry is assumed while truncating the time signal, no discontinuity and related artifacts are introduced in DCT.Two related transforms are the discrete sine transforms (DST) which is equivalent to a DFT of real and odd functions, and the modified discrete cosine transforms (MDCT), which is based on a DCT of overlapping data.

4.6 DISCRETE SINE TRANSFORM DST is identical to Shifted Discrete Fourier transform spectra of signals that are certain permutation modifications of original signal. The DST is defined by the following equation

……………………………….. 4.13 It also has a property of energy compaction that can be used for audio watermarking the value of inverse sine transform can be calculated by the following equation

………………………………. 4.14 Where n=0…………N-1;4.7 DIFFERENCE BETWEEN THE DCT AND THE DST By applying the DCT to 8 identical values (x (i) = 100, i=1, 2 …., 8) it can see that the DCT compacts all the energy of the data into the single DC coefficient whose value is identical to the values of the data items. Applying the DST to the same eight values, on the other hand, results in seven AC coefficients whose sum is a wave function that passes through the eight data points but oscillates between the points [14].So it can say that the DST has compacted the energy input values by using more AC coefficients than the DCT. This example, with the fact that the DCT produces highly décor related coefficients, argues strongly in favor of using the DCT in data compression in contrast to DST.

CHAPTER-5 DIFFERENT ALGORITHM OF AUDIO

WATERMARKING

A Water mark can be added with an audio signal either in time domain or in frequency domain. the commonly used time domain algorithms are LSB Coding, Echo hiding, patchwork algorithms The main advantage of time-domain audio watermarking is that it is relatively easy to implement, and requires few

computing resources, low cost, fast speed . The main disadvantage of time domain algorithm is that it is weak against various signal processing attacks such as compression and filtering. To embed watermark in time domain we need to reshape the watermark in order to maintain a good perceptual quality .if we add the watermark without reshaping it according to an audio signal then it may produce an audible distortion.Frequency Domain algorithms are Phase coding and coefficient quantization embedding the watermark by modifying the coefficients of the transform outputs Frequency domain watermarking is very robust again various signal processing attacks again it maintains good perceptual quality the main disadvantage of frequency domain watermarking is that it is somewhere more complex as compared to the time domain watermarking and it takes more time to implement watermarking in time domain The common transformation techniques which are used to convert time domain signal in to frequency domain are[1] Discrete Fourier Transform[2] Discrete cosine transforms[3] Short Time Fourier transforms[4] Discrete wavelet transforms

Then there are lots of algorithms with the help of which we can embed water mark in an audio signal and there are some commonly used algorithms.

ALGORITHM 1:-(a) Watermark embedding algorithm[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Framing the sample’s of audio signal so that easily we can embed watermark the length of frame are usually 2048 sample per frames (it is not necessary to take 2048 samples in each frame we can choose any frame size according to our requirement)[3] Take discrete Fourier transform (D.F.T.) of each frame [4] Embed watermark with the prominent peaks of the real coefficient of discrete Fourier transform .Embedding watermark means modifying the prominent peaks of audio signal according to the watermark signal such that to maintain high degree of robustness and good perceptual quality(although there is a tradeoff between these two parameters.

ViW = Vi (1+ α Xi

) ......................................... (1)Where Viw = adjusted magnitude coefficient (watermarked magnitude coefficients) Vi = magnitude coefficient (real DFT coefficients) into which watermark to be embedded Xi = watermark to be embedded with Vi

α = scaling factor decide the strength of watermarking

[5] Take inverse discrete Fourier transform (IDFT) to obtain watermarked audio signal[6] in above algorithm step-2 and step-3 are interchangeable

(B) Watermark Extraction Algorithm

[1] The watermarked audio signal is converted into the frames (the length of frames are 2048 as were used during watermark embedding)[2] Take the discrete Fourier transform of each frame[3] Find out the prominent peaks of real coefficients using peak detection algorithm[4] Extract watermark from the watermarked audio signal with the help of following equation Xi

* = [ (ViW/ Vi )- 1] / α .......(2)

ALGORITHM-2[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Framing the sample’s of audio signal so that easily we can embed watermark the length of frame are usually 2048 sample per frames (it is not necessary to take 2048 samples in each frame we can choose any frame size according to our requirement)

[3] Take discrete cosine transform(D.C.T.) of each Frame

[4] Embed watermark to the prominent peaks of the coefficient of discrete cosine

transform .Embedding watermark means modifying the prominent peaks of transformed coefficients according to the watermark signal such that to maintain high degree of robustness and good perceptual quality(although there is a tradeoff between these two parameters.[5] Take inverse Discrete Transform (IDCT) to obtain watermarked audio signal

(B) Watermark Extraction Algorithm[1] The watermarked audio signal is converted into the frames (the length of frames are 2048 as were used during watermark embedding)[2] Take the discrete Cosine transform of each frame[3] Find out the prominent peaks of coefficients using peak detection algorithm[4] Extract watermark from the watermarked audio signal with the help of following equation Xi

* = [ (ViW/ Vi )- 1] /

α .................................(2)

ALGORITHM-3EMBEDDING ALGORITHM

[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Take Discreate Wavelet Transform of the signal up to certain level (usually 3rd level DWT coefficients gives good result)

[3] Choose appropriate band for embedding watermarking each band of frequencies has certain characteristics and their own set of advantage and disadvantages[4] Frame the wavelet coefficient of that particular band [5] embed watermark to the D.W.T. coefficient of that particular band .Embedding watermark means modifying the prominent peaks of transformed coefficients according to the watermark signal such that to maintain high degree of robustness and good perceptual quality(although there is a tradeoff between these two parameters)[6] Take Inverse Discrete Wavelet Transform [IDWT] to obtain watermarked audio signal(B) Watermark Extraction Algorithm[1] Take Discrete Wavelet Transform of the signal up to the same level (that were used for watermark embedding)[2] Choose the same band of frequencies that were used at the time of embedding watermark.[3] Frame the wavelet coefficient of that particular band [4] Extract the watermark from the D.W.T. coefficient of that particular band

ALGORITHM 4:-WATERMARK EMBEDDING ALGORITHM:

[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Take Discrete Wavelet Transform of the signal up to certain level (usually 3rd level DWT coefficients gives good result)[3] Choose appropriate band for embedding watermarking each band of frequencies has certain characteristics and their own set of advantage and disadvantages[4] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6] Add watermark to the prominent peaks of the D.C.T. Coefficients[7] Take IDCT[8] Combine all frames [9] Take IDCT to obtained watermarked audio signalWATERMARK EXTRACTION ALGORITHM:[1] Take Discrete Wavelet Transform of the watermarked signal up to the same level (that were used during embedding)

[2] Choose the same band of frequencies for embedding watermark.[3] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6]Find the prominent peaks of the D.C.T. Coefficients with the help of peak detection algorithm[7] Extract watermark from these prominent peaks

ALGORITHM 5:- Watermark Embedding Algorithm:

[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)

[2] Take Discrete Wavelet Transform of the signal up to certain level (usually 3rd level DWT coefficients gives good result)[3] Choose low frequency band(approximate coefficients) for embedding watermarking the lower frequency band contain most of the information of the signal hence it offer high degree of robustness against various signal processing attack but as we increase the degree of Watermarking(degree of α) it degrade the perceptual quality. So we need to choose the value of α very carefully.

[4] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6] Add watermark to the prominent peaks of the D.C.T. Coefficients[7] Take IDCT[8] Combine all frames [9] Take IDWT to obtained watermarked audio signalWatermark extraction algorithm[2] Take Discreate Wavelet Transform of the signal up to the same level (that were used during watermark embedding)[3] Choose same low frequency band (approximate coefficients) that were used during watermark embedding process.[4] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6] Find the prominent peaks of the D.C.T. Coefficients with the help of peak detection algorithm[7] Extract watermark

ALGORITHM 6:- Watermark embedding algorithm:[1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Take Discrete Wavelet Transform of the signal up to certain level (usually 3rd level DWT coefficients gives good result)[3] Choose high frequency band (Detailed coefficients) for embedding watermarking. the higher frequency band contain very less information of the signal hence it offer high

Degree of perceptibility.ie it maintained a high degree of perceptual quality even for the high value of α but it is less robust against various signals processing attack [4] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6] Add watermark to the prominent peaks of the D.C.T. Coefficients[7] Take IDCT

[8] Combine all frames [9] Take IDWT to obtained watermarked audio signalWatermark extraction algorithm:[1] Take Discreate Wavelet Transform of the signal up to the same level (that were used during transmission)[2] Choose the same high frequency bands (Detailed coefficients) that were used during embedding watermark. [3] Frame the wavelet coefficient of that particular band [4] Take the discrete cosine transform of the frames of that particular band.[5] Find the prominent peaks of the D.C.T. Coefficients using peak detection algorithm[6] Extract watermark

ALGORITHM 7:- [1]The audio signal is first convert into digital signals (sampled, quantized, and then analog to digital coding)[2] Take Discreate Wavelet Transform of the signal up to certain level (usually 3rd level DWT coefficients gives good result)[3] Simultaneously choose high frequency band (Detailed coefficients) as well as Low frequency band (approximate coefficients) for embedding watermarking. The higher frequency band [Detailed Coefficients] contains very less information of the signal hence it maintained a high degree of perceptual quality even for the high value of α but it is less robust against various signals processing attack. The lower frequency band [Approximate Coefficients] contains most of the information of the signal. Hence it offer high degree of robustness even for the low value of α but adding watermark at these frequency my effect the perceptual quality of the signalSo we embed the watermark at higher frequency for high value of α (for example α =0.3) whereas at lower frequency for the low value of α (for example α = 0.1)

[4] Frame the wavelet coefficient of that particular band [5] Take the discrete cosine transform of the frames of that particular band.[6] Add watermark to the prominent peaks of the D.C.T. Coefficients[7] Take IDCT[8] Combine all frames [9] Take IDWT to obtained watermarked audio signal

Watermark extraction algorithm [2] Take Discreate Wavelet Transform of the watermarked audio signal up to the same level that were used for watermark embedding [3] Simultaneously choose the same high frequency band (Detailed coefficients) as well as same Low frequency band (approximate coefficients) that were used during watermark embedding. [4] Frame the wavelet coefficient of those particular bands (high frequency bands as well as low frequency bands) [5] Take the discrete cosine transform of the frames of those particular bands. (High frequency bands as well as low frequency bands)[6Find out the prominent peak of the D.C.T. Coefficients of that particular bands(high frequency bands as well as low frequency bands) with the help of peak detection algorithm[7] Extract watermark using watermark extraction algorithmComparison of different algorithm of an audio watermarkingWhiling embedding watermark with an audio signal the most important thing is that where we are adding watermark (with which frequencies band)

- If we add watermark with very higher frequency band then it will offer a very good

perceptual quality but somewhere poor Robustness. Even if we add a large amount of information with higher frequencies then the perceptual quality of original audio signal will not be affected. Hence embedding watermark with higher frequencies can carry large amount of information with it.

- If we add watermark with lower frequency band then it will offer high degree of robustness against various intentional and unintentional signal processing attacks but poor perceptual quality. As we increase the strength of watermark (or degree of watermark i.e. by increasing the value of α) the robustness of watermarking increases but the perceptual quality of the signal decreases. So with lower frequency components we cannot maintain the value of α very high.

- Frequency band in between the higher and lower frequency is called middle frequency band if we choose the middle frequency band for embedding watermark information then it is a compromise between lower and higher frequency band i.e. middle frequency band offer robustness greater than lower frequency band but higher than higher frequency band. Similarly middle frequency band offer perceptual quality lower than higher frequency band but higher than lower frequency band.

- We can also simultaneously choose lower and higher frequency bands for adding watermark

information but while adding a watermark with lower frequency components than the value of α should be low. While adding watermark with higher frequency components than the value of α may be high.

- While embedding watermark with an audio signal then most important thing is that which frequency components are selected for adding a watermark

- If we add watermark with low frequency components then it will offer a high degree of robustness against intentional or unintentional signal processing attacks Like Low pass filtering, compression, adding AWGN, Rs- sampling, re quantization Cropping etc. But with low frequency component we cannot increase the strength or degree of watermark otherwise it will affect the perceptual quality of original audio signals.

- If we are adding watermark with higher frequency components of audio signals then it will offer a very good perceptual quality but it will not provide good robustness against common signal processing attacks

- So the selection of frequency band is very important whether we are choosing lower frequency band, middle frequency band or higher frequency band for embedding the watermark information.

- Every watermarking algorithm has its unique set of application advantages and disadvantages some watermark algorithm is suitable for application like information carrying (Watermark algorithms which have high capacity). In which watermark is added with higher frequencies band. On the other hand some watermarking algorithm are suitable for applications where security is important in which watermark is added with lower frequency components of an audio signal.

Advantage of using wavelet Transform: - While embedding the watermark with an audio signal then most important decision is that where to embed watermark with an audio signal i.e. with which frequency bands, with what amplitudes and with what times. Because the three main requirements of audio watermarking are Robustness, Imperceptibility, and capacity are much depends on the decision that where you are embedding the watermark information.If we choose lower frequency band for embedding watermark information then it will give you high degree of robustness against various common signal processing attacks like Re sampling, Re quantization, Low pass filtering, Compression, Cropping etc. But the perceptual quality (Imperceptibility) of original audio signal degraded. Again we cannot add more information with these frequency bands again the strength of

watermarking (or degree of watermarking α) should be maintained low otherwise it will affect the perceptual quality of the signal.If we choose higher frequency coefficients for embedding watermark information then we can add more information along with original audio signal again we can maintain the degree of watermarking or strength of watermarking very high so embedding watermark information with high frequency coefficients results in very high capacity and high perceptual quality but at the same time the robustness of watermark will be less against various intentional or unintentional signal processing attacks.So we can conclude that the Robustness, Imperceptibility and Capacity are three main requirement of watermarking and there is a Tradeoff between these three requirements of watermarking. If we try to improve one requirement then we need to compromise with another requirement. Again there are Different sets of advantages, disadvantages and applications of different technologies. In certain applications Robustness is important whereas for some other applications imperceptibility or capacity is required.

The main advantage of using wavelet transform while embedding watermark information with an audio signal is that wavelet transform separate the audio signal in to the band of frequencies.

What exact is going on in the process of taking wavelet transform we passes the original audio signal into the series of Low pass filter and high pass filter the output of high pass filter is called detailed coefficient while the output of low pass filter is called approximate coefficient. These approximate coefficients are against passed with high pass filter and low pass filter and again we get a new set of detailed coefficient and an approximate coefficient these approximate coefficients are again passed through the high pass filter and low pass filter and again we get the new set of detailed coefficient and approximate coefficients and so onSo after taking the wavelet transform of any signal we will have the different frequency coefficient present in that signal what are the amplitude of these frequency components and at time axis where these frequency coefficients are existing. In the other world we can say that after taking the wavelet transform we will have a three dimensional graph between Time, Frequency and amplitude so it is very easy to take the decision that where( at which time, which frequencies , and with which amplitudes) to add watermark signal. CHAPTER6 CHECK THE PERFORMANCE OF AN ALGORITHM AND EXPERIMENTAL SETUPThe algorithm we are going to Propose takes advantages of both D.W.T. and D.C.T. The main advantage of D.W.T. is multi resolution Property

and better spectral localization whereas the main advantage of D.C.T. is its energy compressing property i.e. ability in compressing energy of the signal in few coefficients Audio watermarking algorithm can be divided into the parts

[1] Watermark Preprocessing

[2] Watermark Embedding Algorithm

[3] Watermark Extraction Algorithm 6.1 Watermark preprocessing:

A watermark (gray scale image) cannot be directly added to the audio signal first we need to process (gray image) our watermarkWatermark preprocessing has the following steps

STEP 1: convert the gray-scale image watermark image into Two dimensional matrix whose size is M*N

STEP 2: Convert gray scale watermark image into binary image

STEP3: with the help of Arnold transform we can scramble the watermark image and get scrambled image.

STEP 4 Convert Two dimensional image matrix into one dimensional image vector W of length M*N

5.2 Watermark embedding algorithm

STEP1: First we decompose the original audio signal with the discrete wavelet transform up to 3 level. After the wavelet decomposition we get the approximate coefficient A3 and Detailed coefficient D1, D2and D3.

STEP2: Select the low frequency coefficient of decomposed signal A3 and this approximate coefficient are then converted into non overlapping frames.

STEP3: apply D.C.T. to each frame and calculate D.C.T. coefficients.

STEP 4: With the help of peak detection algorithm the prominent peaks are fined and then watermark is embedded into selected N peaks using the following equation.ViW = Vi (1+ α. Xi) ………………………………………………………………… [5.1]Where Viw = watermarked magnitude coefficients Vi = magnitude coefficient into which watermark to be embeddedXi = watermark to be embedded with Vi

α = scaling factor decide the strength of watermarking STEP 5: Apply the inverse discrete cosine transform to each frames

STEP 6: Combine each frames

STEP 7: Applied IDWT for getting watermarked audio signal

6.3 Watermark extraction algorithm

STEP1: First we decompose the Watermarked audio signal with the discrete wavelet transform up

to 3 levels. After the wavelet decomposition we get the approximate coefficient's A3 and Detailed coefficient's D1, D2, D3.

STEP2: Select the low frequency coefficients of decomposed signal A3 and this approximate coefficient are then converted into non overlapping frames.

STEP3: Apply D.C.T. to each frame and calculate D.C.T. coefficients.

STEP4: Extract the highest prominent peaks from the D.C.T. coefficients.Which are located at the same position in the embedding process with the help of peak detection algorithm.

STEP 6: With the help of watermark extraction algorithm the watermark vector can be calculated with the help of following equation

Xi* = { ( ViW/ Vi) – 1} / α ………………………………………………………[5.2]

STEP7: Convert extracted watermark vector back into scrambled image

STEP 8: The original image of binary format can be obtained by Appling anti Arnold transform .

6.4 Experimental Setup

For checking the performance of audio watermarking five different type of 16 bit MONO audio signals S1,S2,S3,S4 and S5 are considered which are sampled at the rate of 44.1 KHZ.15*15 gray scale image is used as a watermark . The size of watermark is depend on the length of audio signal. Matlab is used for the simulation of algorithm and for obtaining the ResultsPerceptibility is measured in terms of signal to noise ratio and listening test , using the ITU-R BS.1284 standardized 5-point grading scale Robustness can be measured in terms of normal correlation

6.5 ResultsTo measure imperceptibility, we use Signal-to-Noise ratio (SNR) as an objective measure. Imperceptibility is related to the perceptual quality of the watermarked signal. It ensures that the quality of the signal is not perceptible to a listener.

Type of SignalS.N.R.(α =0.1)

S.N.R.(α =0.2)

S.N.R.(α =0.3)

S1 26.06 22.16 19.40S2 26.29 22.23 20.10S3 26.93 23.06 20.96S4 27.12 23.10 21.43S5 29.44 24.34 21.90

Table 6.1 Signal-to-Noise ratio (SNR)

The subjective measure of imperceptibility is done by listening test Average ITU-R grade value of the proposed watermarking system is shown in table 3

Table 6.2 ITU-R Grade for Imperceptibility

Robustness Test Watermark should be robust again common signal processing attack. Robustness is measured in terms of the Normal Correlation (N.R.).Various common signal-processing attacks are used to check the robustness of the proposed scheme are given below (A)MP3 Compression 64 KBPS: The MPEG-1 layer-3 compression is applied. The watermarked audio signal is compressed at the bit rate of 64

Type of signal

ITU-R Grade(α =0.1)

ITU-R Grade(α =0.2)

ITU-R Grade(α =0.3)

S1 5 5 5S2 5 5 5S3 5 5 5S4 5 5 5S5 5 5 5

kbps and then decompressed back to the WAVE format (B) Re sampling: The watermarked signal, originally sampled at 44.1 kHz, is resample at 22.05 kHz, and then restored back by sampling again at 44.1 kHz.(C) Re quantization: The 16-bit watermarked audio signal is re-quantized down to 8 bits/sample and then back to 16 bits/sample.(D) Additive white Gaussian noise (AWGN): White Gaussian noise is added to the watermarked signal until the resulting signal has an SNR of 15 dB. (E) Low Pass Filtering: A second-order Butterworth filter with cut-off frequency 4 kHz is used. (F) Cropping: 10 % segments are removed from the watermarked audio signal at the beginning and subsequently replaced by segments of the original signal.

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 1 1 1S4 .978 1 1

S5 .997 1 1

Table 6.3 Normalized Correlations after M.P.-3 Compression

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 1 1 1S4 .988 1 1S5 1 1 1

Table 6.4 Normalized Correlation after Re sampling

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 1 1 1S4 1 1 1S5 1 1 1

Table 6.5 Normalized Correlation after Re quantization

Table 6.6 Normalized Correlations after AWGN

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 .998 1 1S4 1 1 1S5 .967 1 1

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 1 1 1S4 1 1 1S5 .998 1 1

Table 6.7 Normalized Correlation after cropping

SIGNAL TYPE

N.C.( α =0.1)

N.C.( α =0.2)

N.C.( α =0.3)

S1 1 1 1S2 1 1 1S3 1 1 1S4 .966 1 1S5 .989 1 1

Table 6.8 Normalized Correlation after Low pass filtering

6.6 Result analysis The simulation result of proposed algorithm for imperceptibility requirement is very good according to IFPI (International Federation of the Phonographic Industry) standard the S.N.R. of watermarked audio signal should be above 20 db. In our proposed algorithm for all the signals and for different value of α the SNR has always been above 25 db refer to table no 2 of chapter 6 which shows good perceptual quality of watermarked signal.

The subjective measure of imperceptibility is done by listening test Average ITU-R grade value of the proposed watermarking system Is shown in table 3

which is always 5 shows excellent perceptual quality of watermarked signal.

Again the watermark is very robust again various type of signal processing attacks the value of normal correlation is always 1 for the value of α=0.2 and α=0.3 for all the signals. Refer table no 4 to table no 9 of chapter 6 even for the value of α=0.1 it is again one for most of attacks and for few attacks (compression and low pass filtering) it is very close to one . Refer table no 4 to 6 of chapter 6. The simulation results shown above in the table prove that the proposed method is quite robust again various signal processing attacks as well as maintain very good perceptual quality.