Acoustic analysis of deaf speech using digital processing techniques

The development of speech-training aids for the deaf requires an understanding of how the acoustic characteristics of deaf speech differ from those of normal speech. The analysis of deaf speech presents problems in that formants may be closely spaced relative to their bandwidths, or unusual variations in voicing may make the separation of source and vocaltract characteristics more difficult than for normal speech. The fundamental frequency contours of deaf children exhibit unusual characteristics, and a second major problem of interest is the quantitative specification of these contours and how they differ from those of normal children. Two digital-processing techniques which are well suited for these problems are the chirp z transform and the short-term orthogonal polynomial analysis. Application of these techniques in the acoustic analysis of the speech of deaf children is discussed.