Skip to main content Accessibility help
  • Print publication year: 2010
  • Online publication date: June 2012

16 - Source coding I: speech and audio coding



Source coding or data compression is used to remove redundancy in a message so as to maximize the storage and transmission of information. In Chapter 14, we have introduced Shannon's source-coding and rate-distortion theorems. We have also introduced lossless data compression based on the source-coding theorem. In Chapters 16 and 17, we will address speech/audio and image/video coding. Lossy data compression is obtained by quantizing the analog signals, and the performance of quantization is characterized by the rate-distortion bound. Source coding is the procedure used to convert an analog or digital signal into a bitstream; both quantization and noiseless data compression may be part of source coding.

Coding for analog sources

Source coding can be either lossy or lossless. For discrete sources, a lossless coding technique such as entropy coding is used. Huffman coding is a popular entropy coding scheme. Lossless coding uses more radio spectrum. For analog sources, lossy coding techniques are usually used.

PCM is a digital representation of an analog signal. The signal magnitude, sampled regularly at uniform intervals, is quantized to a series of symbols in a digital, usually binary code. This can be performed by using A/D converters. The demodulation of PCM signals can be performed by DACs. The PCM code is the original waveform for source coding. There are three approaches to source coding.