An Improved Spectral Subtraction Algorithm for Noise Reduction in Cochlear Implants with Increasing Number of Channels

Cochlear implants are widely accepted as the unique and most effective way for individuals with severe to profound hearing loss to restore some degree of hearing. Speech enhancement strategies play an extremely important role in optimizing the cochlear implant. In this study, a noise reduction algorithm was proposed for cochlear implant. To improve the performance of cochlear implants in noisy environments, a noise reduction algorithm was proposed for cochlear implant that applied a spectral subtraction using the classifications between the speech and the noise dominants in each channel. The proposed classifications use the standard deviation of the spectrum of observation signal in each channel. The performance of the proposed noise reduction algorithm was evaluated by segmental signal-to-noise ratio (SNR) using Noisy92 sentences embedded in babble, car noise and train at 0–20 dB. SNR and subjective listening tests were assessed with 15 normal hearing listeners using a specific cochlear implant (CI) simulator in Clinical audiology of Isfahan University of Medical Sciences, Iran, in June 2012. In addition, to evaluate the effect of the channel on speech recognition, 6, 8 and 16 channels of filter banks were used. At all SNR values, subjects performed better with proposed enhanced algorithm by 45 percentage points. Based on comparing segmental SNR with spectral subtraction, and visually inspecting the enhanced spectrograms and subjective listening test, the proposed method was found to effectively reduce noise while minimizing distortion to speech.