Application of Fractional Calculus in Speech Enhancement: A Novel Approach

In day-to-day life, voice is most common part of communication. While the processing of speech signal in data science, mostly noise affects these signals. As a preprocessing technique in data mining, the signal is to be normalized and noise free that can process fast. In this work, the concept of fractional calculus is applied for enhancement. For this Grunwald-Letnikove (G-L), Caputo and Riemann-Liouville (R-L) methods are used for enhancement by decomposing the signal. The comparison results show that R-L outperforms better than other methods in terms of signal-to-noise-ratio (SNR) perceptual evaluation of speech quality (PESQ) and log spectral distance (LSD).