MPEG-2 AAC - Advanced Audio Coding

factor of 2 better than MPEG-1 Layer 2
factor of 1.4 better than MPEG-1 Layer 3

Non-Backwards Compatible (NBC)

Features

• expanded bit rates: 8-64+ Kb/s/channel
• expanded sample rates: 8-96kHz (better than CD quality is possible)
• multichannel sound: 48 main audio, 16 LF effects, 16 overdub, 16 assoc. data, 16 programs
• profiles: main, low complexity, scalable sampling rate

Techniques - encoder/decoder

1. Filter bank and block(window) switching: MDCT
• 1024 coef. long block for coding gain
• 128 coef. short block for time resolution
• start/stop blocks
2. Temporal Noise Shaping: distributes quantization noise in time using prediction
3. Prediction: uses LPC speech coding technique for appropriate signals
4. Joint Coding: removes stereo redundancy M/S (L+R/S-D), or R/L joint intensity
5. Scalefactors: shared gain value for a band of spectral coefficients to shape noise (49 bands for long window)
6. Quantization: finer control of coef. quantization resolution (4/3 power)
7. Noiseless Coding: huffman coding removes redundancy in banded and sectioned quantized spectrum (11 codebooks)

Major contributors

• FhG - Fraunhofer IIS (Institute for Integrated Circuits)
• AT&T
• Dolby Laboratories
• Sony