Time-scale modification of speech signals for language-learning impaired children

Speech signals can be modified to use as part of a program for training language-learning impaired children. The intent was to provide a better training source for recognition of fast acoustic signals that are embedded in speech and unambiguous perceptions of speech. For this purpose, speech signals were slowed by a factor of two without altering the pitch and speaker dependent features of the recorded voice to provide more easily understood speech. Various audio recordings of speech material were processed to evaluate the performance of three different time-scale modification (TSM) algorithms. It has been demonstrated that the “waveform similarity overlap-and-add (WSOLA)” algorithm increases the quality of the time-scale expanded speech over the other well-known TSM algorithms and produces high quality speech with the desired time-scale