Adaptive-rate sampling applied to the efficient encoding of speech waveforms

Considers a procedure for spectral manipulation, based upon a time domain sampling algorithm, applied to signals having spectra comprising multiple (>or=2) separated band-pass elements. In essence, the algorithm provides an analytical and practical tool to enable spectral manipulation to be carried out in a systematic manner. This allows nonuniform adaptive sampling rates to be applied to a multiple band-pass analogue signal if the band structure is first characterised. Also, time variation in the analogue signal structure can be tracked. The paper describes the results of experimental work involving the adaptive sampling and Hadamard transform coding in an adaptive-rate sampling system and subsequent recovery of speech-type signals with a multiple band-pass structure.