Basic Audio Compression Techniques

In this chapter, compression of audio information is reviewed, with special consideration paid to speech compression. To begin with, we recall some of the issues covered in Chap. 6 on digital audio in multimedia. Here, this is combined with techniques that exploit the temporal redundancy present in audio signals. We extend the Pulse Code Modulation (PCM) scheme to DPCM, prepending the word “Differential,” as briefly introduced in Chap. 6 but fleshed out here. Specifically, in this chapter, we look at ADPCM, Vocoders, and more general Speech Compression: LPC, CELP, MBE, and MELP. Adaptive DPCM is ADPCM. In speech coding, a number of standards have evolved and we set these out here, including some of their fundamental strategies. We then go on to study coders (encoding/decoding algorithms) specifically aimed at speech compression. The properties of Vocoders are examined, including the notion of phase insensitivity, channels, and formants. Next, LPC (Linear Predictive Coding) vocoders are discussed, followed by CELP (Code Excited Linear Prediction), a more complex family of coders. Hybrid Excitation Vocoders are another large class of speech coders, and we round the discussion off by having a look at MBE (Multi-Band Excitation) and MELP (Multiband Excitation Linear Predictive) vocoders.