Speech Enhancement Based On Dual Tree Complex Wavelet Transform

Abstract

The dual tree complex wavelet transform (DT-CWT) is an efficient tool for many signal processing applications (i.e. image coding, image fusion, image enhancement, pattern recognition and image resolution enhancement etc.),  since it has advantages of near shift-invariance, directional selectivity for two or more dimensions and low computational complexity. In his paper, a new speech enhancement method based on the DT-CWT is proposed in order to exploit the aforementioned advantages of the DT-CWT in speech enhancement. An efficient estimator, multiplicatively modified log-spectral amplitude (MM-LSA) estimator, is used for the enhancement of noisy subband wavelet coefficients. The objective tests (SNR, segmental SNR and PESQ-MOS) are used to evaluate the performance of the proposed speech enhancement method. The performance results of the proposed method are compared with those of the other well-known wavelet-based methods. The objective test results and experimental results show that the DT-CWT outperforms the standard wavelet-based methods in speech enhancement.

Keywords: Speech Enhancement, Dual Tree Complex Wavelet Transform, Multiplicatively Modified Log-Spectral Amplitude Estimator.