Speech analysis using wavelet transforms dedicated to cochlear prosthesis stimulation strategy

In this article, we studied speech analysis using wavelet transforms in order to conceive a flexible stimulation strategy for cochlear prostheses. In our first experiment, we used two types of wavelets that are convenient for such speech analysis. This was based on parameterisation techniques of word by using bank filtering methodology. However, the constructed filters had the considered wavelets´ shape impulse response. In fact, as a first experimentation, Morlet wavelet shape was tested and compared to gammachirp wavelet shape. Two filter bank models were then conceived according to these two wavelet shapes in order to apply numerical filtering. Our purpose is to exploit such filtering for constructing a completely flexible cochlea´s stimulation strategy. So, this would be very useful for specifying performances of each of these two analyses and for providing comparisons that would be useful for an ulterior strategy implementation on the prosthesis external part.