Authors : Haider J. Abd, Hadeel Abdallah and Ali Shabban
Abstract: Speech signals play an important role in digital signal processing. When this signal passes through the medium, it interacts with the noise, so, noise must be removed without affecting the original signal. Denoising methods are a compromise between removing as much noise as possible and maintaining signal integrity. In this work, a Hybrid Bacterial Foraging Particle Swarm Optimization Model (HBFPSO) was proposed to estimate the threshold value without any prior information on signal and noise distribution to measure kurtosis function of remaining noise to locate optimal value of threshold when kurtosis value is maximized. It is noted that the suggested denoising technique showed an excellent performance over single models (PSO and BFO) at the same conditions. Furthermore, denoising results showed that the proposed HBFPSO algorithm provided the least MSE which resulted in an improvement of (0.004) compared with (BFO, PSO) algorithms.
Haider J. Abd, Hadeel Abdallah and Ali Shabban, 2018. A New Speech Denoising Technique Based on Wavelet Thresholding and Hybrid Algorithm. Journal of Engineering and Applied Sciences, 13: 10923-10931.