Effect of Intensity Resolution on Wiener Filtering and Spectral Subtraction Method for Noise Suppression in Cochlear Implants

Abstract: Cochlear implants are biomedical devices that serve as advanced hearing aids for profoundly deaf people. In quiet listening conditions and when no background noise is present, cochlear implant devices produce high speech recognition. However, speech recognition using cochlear implant devices significantly drops in the presence of ambient noise. Since speech recognition using cochlear implants drops due to noise, incorporation of noise suppression methods to suppress noise can improve speech recognition using cochlear implants. In this research paper, the speech recognition improvement using two such noise suppression methods namely, spectral subtraction and Wiener filtering was evaluated using cochlear implant simulations. The performance evaluation of the two noise suppression methods was done in a systematic manner by varying the number of intensity steps. In this experiment the intensity resolution for cochlear implant simulation was varied in four different ways, and the performance of Wiener filtering and spectral subtraction was evaluated for the four different values of intensity resolution. Results indicated that the speech recognition using Wiener filtering was significantly higher than the speech recognition obtained by spectral subtraction method for the four values of intensity resolution.

Keywords: Noise Suppression, Cochlear Implants, Intensity Resolution, Spectral Subtraction, Wiener Filter, Spectral Subtraction Method.