Subsampling of a DPCM Speech Channel to Provide Two 'Self-Contained' Half-Rate Channels

The channel-splitting problem described in the abstract is redefined in Fig. 1. The receiving end of a speech communication system is supposed to operate in either a full-rate or half-rate mode depending on whether it has available to it both or only one of the speech subchannels. Respective qualities of speech reproduction are denoted by Qf(B) and Qh(B/2). The nonavailability of one subchannel is a good model for certain types of transmission failure, examples of which are signal fading in mobile radio and speech segment losses in packet switching. With appropriate forms of diversity reception, the second subchannel will be available with probability close to unity when the first subchannel is not. The channel-splitting problem has been re501 SPEECH Q U A L I T Y QF(B) Fig. 1--Definition of the channel-splitting problem. When both of two half-rate components in the transmitted sequence are available at a receiver, conventional fullrate receptions result, with speech quality QF(B). When only one of the components is present, a half-rate receiver recovers an approximation to the full-rate speech, with quality QH(B/2).