5th International Conference on Spoken Language Processing
A subtractive deconvolution algorithm is described which allows one to separate a voiced speech signal into two components, representing the time-invariant and dynamic parts of the signal respectively. The resulting dynamic component can be encoded at a lower data rate than can the original speech signal. Results are presented which validate the utility of decomposing the speech waveform into these two components, and demonstrate the ability of the algorithm to represent speech signals at a reduced data rate.
Bibliographic reference. Thorpe, C. William (1998): "Separation of speech source and filter by time-domain deconvolution", In ICSLP-1998, paper 0244.