DIGITAL AUDIO WATERMARKING ALGORITHM BASED ON NEURAL NETWORKS

• 1. Presenter : behzad ghorbani

2. outline Introduction Digital watermark Principle and embedding Wavelet Transform Watermark embedding Experimental results Conclusions 3. Introduction Text/Image/Audio watermarking refers to embeddingwatermarks in an text/image/audio in order to protect the image from illegal copying and identify manipulation 4. Applications of Watermarking Rights management Contents management Access/copy control Authentication 5. Digital Watermarking? Allows users to embed SPECIAL PATTERN or SOME DATA intodigital contents without changing its perceptual quality. When data is embedded, it is not written at HEADER PART but embedded directly into digital media itself by changing media contents data Watermarking is a key process for the PROTECTION of copyright ownership of electronic data 6. watermarking watermarking is embedded information into data audio Image video Embedding Techniques Original needed Original not needed 7. Features of Watermarking Invisible/Inaudible Information is embedded without digital content degradation,because of the level of embedding operation is too small for human to notice the change. Inseparable The embedded information can survive after someprocessing, compression and format transformation Unchanging data file size Data size of the media is not changed before and after embeddingoperation because information is embedded directly into the media 8. In /visible Watermark 9. Embedding Techniques Spatial domain Original needed Original not neededFrequency domain Original needed Original not needed 10. Spatial domain Least Significant Bit 11. Frequency Domain Watermarking signal to embed Xx0 , x1 ,..., xN Host signal Vv0 , v1 ,..., v N Frequency components(using DCT) Ff 0 , f1 ,..., f N Embedding fifixi Extraction xififif i = watermared singal 12. Frequency Domain Fourier Transform Short Time Fourier Transform (STFT ) Wavelet Transform 13. Fourier transform Stationary Signal Signals with frequency content unchanged in time All frequency components exist at all times Non-stationary Signal Frequency changes in time One example: the Chirp Signal13 14. Fourier transform FT Only Gives what Frequency Components Exist in the Signal The Time and Frequency Information can not be Seen at the Same Time Time-frequency Representation of the Signal is Needed14 15. STFT To analyze only a small section of the signal at a time -- a techniquecalled Windowing the Signal. 16. STFT Overcomes the preset resolution problem of the STFTby using a variable length window: Use narrower windows at high frequencies for bettertime resolution. Use wider windows at low frequencies for betterfrequency resolution. 17. STFT Wide windows do not provide good localization athigh frequencies.17 18. STFT Use narrower windows at high frequencies18 19. STFT Narrow windows do not provide good localization atlow frequencies.19 20. STFT Use wider windows at low frequencies.20 21. Wavelet Transform Wavelet - Small wave. SinusoidWavelet 22. CWT CWT is defined as in other transforms: Translation parameter, measure of timeScale parameter (measure of frequency)C ( , s)Normalization constant1 sf t tt sdtForward CWT:Scale = 1/FrequencyContinuous wavelet transform of the signal f(t)Mother wavelet (window)The results of the CWT are many wavelet coefficients, which are a function of scale (dilation) and translation (position or shift) parameters. 22 23. DWT Multiple-Level Decomposition (Subband Coding )23 24. Watermark embedding Select watermark Lowering watermark dimensional Rseudorandom sorting watermark data Watermark embedding scheme Watermark embedding scheme Neural networks training Digital watermark extraction 25. Select watermark 26. Lowering watermark dimensional 27. Watermark embedding scheme 28. Watermark embedding scheme 29. Neural networks training the triple-layer feed-forward neural network 30. Neural networks training 31. Digital watermark extraction 32. Digital watermark extraction 33. Experimental results 34. Conclusions which based on Neural Networks and in the WaveletDomain Watermark by utilizing still image and embedded into the significant wavelet coefficients of a digital audio signal weights of the neural networks are relationship between the host digital audio and the watermark