Approach to More Efficient Segmentation and Classification of Speech Sounds
William J. Beninghof, M. Ross · The Journal of the Acoustical Society of America · 1968
The success of lumped-parameter analogs of the vocal tract suggests that speech spectra may be optimally approximated by a rational function. As a step in this direction, speech spectra have been approximated by a linear combination of Gram polynomials and have been found to be more efficient than a linear combination of trigonometric functions. The original spectra (100 samples in frequency) and the polynomial and trigonometric approximations have been represented by points in their respective Hilbert spaces, the distance between successive points being a measure of the dissimilarity of successive spectra. Segment boundaries are indicated where the distance between successive spectra exceeds a threshold. The effectiveness in segmentation of connected utterances using these spectral forms will be compared. Results will be presented and implications made for real-time machine recognition of speech. [This research was supported by NASA Fellowships and the Electrical Engineering Department at Northeastern University.]