Improving the Performance of Noise Reduction in Hearing Aids Based on the Genetic Algorithm

Abstract

The noise reduction (NR) approach is a critical component for enabling hearing aid (HA) users to attain improved speech perception capabilities and to reduce the listening effort under noisy conditions. More specifically, the NR can improve the output signal-to-noise ratio (SNR) level to enhance the speech intelligibility and sound quality for individuals with hearing loss. Recently, a discriminative post-filter (DPF) approach has been developed and the preliminary results indicate that the DPF can further improve the output SNR performance. It is important to specify suitable parameters for the DPF; however, it remains unknown how to set these parameters properly. Therefore, this study mainly focuses on the genetic algorithm (GA) to investigate these parameters of the DPF to further enhance the output SNR performance for HA users. The results of the objective evaluations show that under noisy conditions, the parameters of the DPF investigated by the GA algorithm can maintain the sound quality and yield higher SNR performance as compared to the NR approach used solely. More specifically, the proposed method can enhance the output SNR over the NR approach applied singly by 5.31 dB and 4.05 dB at SNR conditions of 0 dB and 5 dB, respectively. It implies that the proposed approach can help HA users to reduce the listening effort and to hear well even in noisy conditions.