Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2008
Session Number:E4
Session:
Number:E4-5
Adaptive Noise Reduction Filter for Speech Using Cascaded Sandglass-type Neural Network
Hiroki Yoshimura, Tadaaki Shimizu, Toshie Matumura, Masaya Kimoto, Naoki Isu,
pp.-
Publication Date:2008/7/7
Online ISSN:2188-5079
Summary:
In this study, we developed a new adaptive noise reduction filter (CSNNFR: Cascaded Sandglass-type Neural Network Noise Reduction Filter) which is extended from the SNNRF (Sandglass-type Neural Network Noise Reduction Filter). The CSNNRF can perform adaptive noise reduction while capturing dynamic characteristics of the speech signal, owing to the plasticity of the SNN. The CSNNRF is suitable for application to the hearing aid since the ease of hearing the speech signal after noise reduction is more important than the denoising rate. We examined the improvement rate of SN ratio and performed a hearing experiment to evaluate the performance of the CSNNRF as a hearing aid.