Speech processing for cochlear implants with the discrete wavelet transform: feasibility study and performance evaluation.

An innovative approach is investigated for speech processing in cochlear implants (CI). Differently from the traditional filter-bank spectral analysis strategies, the proposed method analyses the speech signal by means of the discrete wavelet transform (DWT). Preliminary tests were conducted in order to compare the WT and the filter-bank analysis methods. Additionally, the intelligibility of the speech processed with the proposed WT strategy was tested on normal hearing people by means of the acoustic simulations and a comparison was made with respect to traditional CI algorithms. Results showed that the WT could be a suitable method for speech coding in CIs. The information loss was minimal and, in speech recognition tests, the WT performance was similar to traditional filter-bank strategies.