Temporal Effects in Speech Analysis and Synthesis
Robert J. Scott · The Journal of the Acoustical Society of America · 1966
Experiments in adjusting the time dimension of natural speech have been carried out with a hybrid computing facility. The distribution of spectral energy for successive time samples was derived by scanning the output of a channel-vocoder analyzer, and the relationship between the analytic and synthetic sampling rate was modified in various ways. Time normalization of utterances was done by both linear and nonlinear methods; the nonlinear, arc-length method may be better for automatic speech recognition. The perceptual effects of time adjustment were investigated in an intelligibility test. A dichotic method for time compression of speech was investigated in which odd-numbered time samples were presented to one ear and even-numbered samples to the other ear. Potential applications for time-adjusted speech and possibilities for further research are suggested.