Feature extraction using information-theoretic learning.

A classification system typically consists of both a feature extractor (preprocessor) and a classifier. These two components can be trained either independently or simultaneously. The former option has an implementation advantage since the extractor need only be trained once for use with any classifier, whereas the latter has an advantage since it can be used to minimize classification error directly.

Self-organized acoustic feature map in detection of fricative-vowel coarticulation.

The self-organizing map, a neural network algorithm of Kohonen, was used for the detection of coarticulatory variation of fricative [s] preceding vowels [a:], [i:], and [u:]. The results were compared with the psychoacoustic classification of the same samples to find out whether the map had extracted perceptually meaningful features of [s]. The map distinguished samples of [s] in front of [u:] from those in front of [a:] or [i:] throughout the fricative duration.

Acoustic pattern recognition of fricative-vowel coarticulation by the self-organizing map.

Word-initial samples of fricative [s] preceding vowels [a:], [ae:], [e:], [i:], [u:], [o:], and [y:] in Finnish words were studied with the self-organizing map. An acoustic map was first calculated from speech samples of women without speech disorders, and then the [s] samples were measured on this map. In all 10 subjects the [s] samples preceding the rounded vowels [u:] and [o:] clearly differed from the samples in front of unrounded [a:], [ae:], [e:], and [i:].

Acoustic pattern recognition of /s/ misarticulation by the self-organizing map.

The [s] samples of 11 women, psychoacoustically classified as acceptable/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 spectral vectors, analyzed with the map, were calculated from short-time FFT spectra at 10-ms intervals.

Dysphonia detected by pattern recognition of spectral composition.

The vowel [a:] in a test word, judged normal or dysphonic, was examined with the Self-Organizing Map; the artificial neural network algorithm of Kohonen. The algorithm produces two-dimensional representations (maps) of speech. Input to the acoustic maps consisted of 15-component spectral vectors calculated at 9.