Fractional Sine Transform Based Single Channel Speech Enhancement Technique

Abstract

In this chapter the effectiveness of using discrete fractional sine transformation (DFrST) based single channel speech enhancement is studied. The DFrST transforms a signal or function into any intermediate domain between time and frequency, according to the rotation of time frequency distribution via angle or order. In this method, the input noisy speech is transformed from spatial domain to spatial frequency domain using DFrST. The enhanced speech is obtained by the application of one dimensional DFrST in a combination of Wiener filter algorithm with two step noise reduction and harmonic regeneration noise reduction techniques. The performance of the DFrST-approach is demonstrated through the simulation on real world noisy speech signals dataset. Experimental results proved that DFrST reduces the noise and improves the quality of speech with improved SNR, Seg-SNR, and SNR-Loss over the conventional Weiner filter and discrete fractional cosine transform (DFrCT) based approaches. In Sects. 5.1 and 5.1.1, the definition, derivation, and properties of DFrST are described. Some of the applications of DFrST are listed in Sect. 5.1.2. The proposed approach for speech signal enhancement using DFrST is presented in Sec. 5.2. Sections 5.3 and 5.4 cite the performance evaluation and experimental result analysis, respectively. In Sect. 5.5, the chapter concludes with observations and conclusions.