Speech Enhancement Based on Stationary Bionic Wavelet Transform and Maximum A Posterior Estimator of Magnitude-Squared Spectrum

Abstract

Numerous efforts have focused on the problem of reducing the impact of noise on the performance of various speech systems such as speech coding, speech recognition, and speaker recognition. These approaches consider alternative speech features, improved speech modeling, or alternative training for acoustic speech models. In this chapter, we have detailed our speech enhancement technique proposed previously in literature. This technique integrates our proposed wavelet transform named stationary bionic wavelet transform (SBWT) and the maximum a posterior estimator of magnitude-squared spectrum (MSS-MAP). The SBWT was proposed for solving the problem of the perfect reconstruction existing with the bionic wavelet transform (BWT). The MSS-MAP estimation was employed for estimating speech in the SBWT domain. The experiments were conducted for different sorts of noise and for many speech signals. This proposed technique is evaluated and compared to other popular speech enhancement techniques such as Wiener filtering and MSS-MAP estimation in frequency domain. This evaluation is performed through the computations of the signal-to-noise ratio (SNR), the segmental SNR (SSNR), the Itakura–Saito distance (ISd), and the perceptual evaluation of speech quality (PESQ). The results obtained from these computations proved the performance of the proposed speech enhancement technique. The latter provided sufficient noise reduction and good intelligibility, without causing considerable signal distortion and musical background noise.KeywordsBionic wavelet transformSpeech enhancementMSS-MAPSBWTSNRSegmental SNRISdPESQ