Dipanjon Halder 1230630 MP3 Player
Jul 16, 2015
Dipanjon Halder
1230630
MP3Player
WHAT IS
MP3 ??
MPEG-1 or MPEG-2 Audio Layer III
MP3, is an audio coding format for
digital audio which uses a form
of lossy data compression.
MP3 Cycle
Audio as Physical
Phenomenon Vibrations of object generate sound
Sound propagates as pressure wave
Ear can sense pressure wave
Audio as Analog
Signal Microphone translates waves into varying
voltage
Speaker converts electrical signal into
pressure wave
Digital Signal Recording
Need process to represent analog signal in binary
Steps for processing working mp3 Measure signal (“sampling”)
Translate into binary (“quantization”)
Store or transmit
Reconstruct signal (“excite filter”)
Sample and Sampling Rate
Sampling rate determines quality of representation
Low-rate sampling fails to capture high frequencies
Nyquist-Shannon sampling theorem
“If a function f(t) contains no frequencies higher than W hertz, it is completely
determined by giving its ordinates at a series of points spaced 1/2 W seconds
apart.”
Quantization Samples have continuous value
“Quantization” assigns discrete value to each
sample
Analog-to-digital (A/D) converter n-bit digital output
n bits have 2n possible values
Digital Representation
Digital representation Encode quantized values in binary
Concatenate binary codes of samples
Add meta-information (can be implied if standard is used)
Playback Digital-to-analog (D/A) converter Generates voltage of sample value
Voltage is held for duration of sample period
Low-pass filter to “smooth out” signal
Signal is amplified and sent to speaker
Aliasing
Difference between original and reconstructed
signal
Almost same to the original signal
Sampling and Quantization
Quality
Sampling rate and quantization levels impact
quality
Parameters
Telephony: pulse code modulation (PCM) ITU-T standard G.711 (sample 8000, quantazation-8 bit)
µ-law in U.S. (14-bit samples)
A-law in Europe (13-bit samples)
Encoded signal: 64 kb/s
CD-quality audio: (PCM) Sampling: 44,100 samples per second
Quantization: 16-bit samples
Encoded signal (stereo): (176.4kB/s)
How Can Reduced the Mp3 size???
Compression
Example for loss-less compression: Huffman coding
Variable-length code
Huffman coding for our example
2-bit symbol frequency:
10 (37%), 01 (30%), 00 (23%), 11 (10%)
New encoding
10→0, 01→10, 11→110, 00→111
Encoded sequence marginally better Only 1 bit (2%) shorter
Better on sequences with more redundancies
101001000110101010011101110000010110110001000000101010011001
00101111000001011010110111111101001101111011111111100010010
Conti…… Lossy compression
Uses perceptual coding
Reduces precision of audio components
less audible to humans
Sound is analyzed in a short windows
Analysis in time domain and frequency domain
Coding exploits masking effects
Simultaneous masking: loud sound masks soft sound
Temporal masking: Loud sound masks following soft sound
Reduces data rate considerably
MP3 uses 128kb/s for CD-quality audio
Typical 5-minute song: 4.8MB
16GB MP3 player: more than 3,000 songs