In order to reduce the noise for speech enhancement, we put forward a speech enhancement system based on the combination of spectral subtraction and binary masking. The spectral subtraction method is used to estimate the noise power spectrum and in the binary masking method is used to separate the target voice from other signals by time frequency decomposition. In this article, we combine spectral subtraction and binary masking method with a bank of Gamma-tone filters to reduce the residual musical noise. In the proposed method, first Spectral subtraction algorithm is used for reduction of noise in noisy speech and then, binary masking method is used for monaural speech separation from musical noise introduced by spectral subtraction. We use Channels 1,2,4,8,16,32,64,128 from a bank of 128 gamma-tone filters for implementing the binary mask and then calculate the output. The evaluations confirmed that the proposed method has obviously improved the clarity and intelligibility of speech signal in SNRs(Signal to Noise Ratio) about 5 to 19 dB.
کلید واژگان :speech enhancement, spectral subtraction, binary masking, Gamma-tone filter
ارزش ریالی : 350000 ریال
با پرداخت الکترونیک