Efficient speech enhancement technique by exploiting the harmonic structure of voiced segments

One of the most significant applications of the speech processing is speech enhancement. Several classical algorithms have been proposed and they suffer from one serious drawback of enhancing speech signal under very low signal to noise ratio. In this paper, this is accomplished by exploiting the harmonic structure of voiced segments. This method improves speech quality by suppressing the noise in the frequency domain with the use of a spectral weighting function. Two design parameters are included in the suppression gain, namely, the frequency-dependent noise flooring parameter (FDNFP) and the gain factor. Enhanced harmonic structures are incorporated into the FDFNP by time-domain processing of the linear prediction residuals of voiced speech. Further enhancement of the harmonics is achieved by adaptive comb filtering with a peak-picking algorithm. Simulation results prove that an admissible level of speech distortion is obtained as comparable to those caused by conventional noise reduction techniques. The performance of the enhancement method was evaluated by the modified bark spectral distance (MBSD), ITU-Perceptual Evaluation of speech Quality (PESQ) scores, composite objective measures and listening tests.