Speech coding via splicing and ADPCM
RE Bogner · 1st IASTED International Symposium on Signal Processing and its Applications · 1987
By removal of some of the nearly repetitive sections of a speech waveform, and replacement of them by previous sections it is possible to reduce the data required for intelligible speech. It is necessary to pay careful attention to the continuity of the waveforms at the instants of cutting and splicing, and also to the durations of the sections to be excised and repeated. Such waveform manipulations can produce time compression if the waveform is spliced together without repetition, and a 2:1 compression is practical. This compressed waveform, which requires only half the data of the original speech, can be expanded again by identifying and repeating suitable segments splice points. Thus two speech channels may be compressed to require the same data capacity as one unmodified channel. Given that existing standards for ADPCM at 32 kilobit/s already halve the 64 kilobit/s standard requirement then combination of the two reduction methods should provide another halving of the data rate requirement.