Speech analysis for hearing impaired using chirped wavelets

Wavelet techniques are utilized to process speech signals for the hearing impaired. We are interested in schemes that isolate and enhance short duration, broadband speech elements (such as unvoiced consonants) while still capturing narrowband speech components. Discerning unvoiced consonants presents an obstacle for hearing impaired persons. A dyadic scheme is proposed, using chirped wavelets, to seize the characteristics of broadband speech elements. The method is tested on speech data, in noise free and noisy conditions, and compared with sinusoidal based wavelets. Results demonstrate that chirped wavelets are useful in the analysis and enhancement of speech for the hearing impaired.