Audio encoding
Audio compression was developed some time before the formation of MPEG and its Eureka 147 project. Eureka 147 developed the MUSICAM (Masking pattern adapted Universal Sub-band Integrated Coding and Multiplexing) jointly with some European manufacturers. MUSICAM was designed for digital audio broadcasting (DAB). In parallel with this, another system was being developed called ASPEC (Adaptive Spectral Perceptual Entropy Coding) jointly by AT&T Bell Labs, Thomson and others. ASPEC included high compression for audio transmission on ISDN lines. Following a comprehensive subjective testing on both sys- tems by the Swedish Broadcasting Corporation, MPEG audio group combined both systems into a common standard with three levels known as layers: I, II and II. The three layers differ in coding complexity and per- formance in terms of bit rate reduction. Layer I is a simplified version of MUSICAM which provides low compression rates at low cost. Layer II employs MUSICAM technology in full; it provides high compression rates and is generally employed in DAB and digital television (DTV). Layer III (commonly known as MP3) combines the best features of both techniques, providing extremely high rates of compression. It is mainly employed in music download and telecommunication applications where high compression ratios are necessary. Further development has produced advanced audio coding (AAC) with increased rates of compres- sion making multi-channel surround sound a practical possibility for home users.
Principles of MPEG-1 audio
The process of audio compression starts with digitising the L and R audio signals before going into the MPEG audio encoder as illustrated in Figure 6.1. Digitising involves sampling the L and R channels separately and then converts the samples into multi-bit pulse-coded modulation (PCM) codes by the quantiser. The output is a series of PCM pulses representing the stereo audio channels. This is followed by MPEG encoding.
Before MPEG encoding, the audio is uncompressed PCM with a bit rate that is dependent on the chosen sampling rate. MPEG audio supports three sampling rates, 32, 44.1 and 48 kHz. At a sampling rate of 48 kHz, the bit rate of an uncompressed PCM audio may be as high as 480 kbps per channel, i.e. almost 1 Mbps for stereo. Audio compression will reduce the bit rate by up to a factor of 7 or 8, depending on the coding layer used. At a sampling frequency of 48 kHz, these are the typical bit rates for hi-fi quality sound:
● Layer I: 192 kbps per channel (384 kbps for stereo)
● Layer II: 128 kbps per channel (256 kbps for stereo)
● Layer III: 64 kbps per channel (128 kbps for stereo)