A Study of Noise Estimation of Forward and Backward MCRA for Speech Enhancement

• Main Authors: Chien-Pei Hung, 洪千焙
• Other Authors: Yeou-Jiunn Chen
• Format: Others
• Language: zh-TW
• Published: 100
• Online Access: http://ndltd.ncl.edu.tw/handle/76319478965330966833

碩士 === 南台科技大學 === 電機工程系 === 99 === Hearing impaired will affect communication abilities, cognition and personal connection, digital hearing aids can help a user with hearing impairment to use residual hearing. The digital hearing aids have high flexibility in fittings, hearing loss compensation modulate easily, and wide dynamic range compression. In the noise environment, speech and noise are be modified simultaneously, that can’t avoid the interference of noise and hinder on communications. Therefore, the noise suppression can effectively improve the performance of digital hearing aids. In this paper, using speech enhancement technique to suppress noise components, the noise is estimated by minima controlled recursive averaging, based on area minimum that have estimate delay problem on rapid increase noise, we propose a backward minima controlled recursive averaging method to increase the performance of non-stationary noise estimation. We adopt minimum mean-square-error based log-spectral amplitude estimator to obtain the gain function. Finally, we use harmonic regeneration approach to restore suppressed speech components. The experimental results show that we proposed method can efficient increase abilities of noise estimation in non-stationary noise environment and to further promote the speech intelligibility.