ACOUSTIC PATTERN-RECOGNITION OF [S] MISARTICULATION BY THE SELF-ORGANIZING MAP

The [s] samples of 11 women, psychoacoustically classified as acceptab le/unacceptable, were studied with the self-organizing map, the neural network algorithm of Kohonen. The measurement map had been previously computed with nondisordered speech samples. Fifteen-component spectra l vectors, analyzed with the map, were calculated from short-time FFT spectra at 10-ms intervals. The degree of audible acceptability correl ated with the location of the sample on the map. Spectral model vector s in different map locations depicted distinguishing spectral features in the [s] samples analyzed. The results demonstrate that self-organi zed maps are suitable for the extraction and measurement of acoustic f eatures underlying psychoacoustic classifications, and for on-line vis ual imaging of speech.