Speech Signal De-noising using Wavelet Transform and Different Standard Softwares: Performance Evaluation and Comparisons Study

Speech signal processing is used in many applications. There exist situations where conversation takes place under high noise environment. A specific example is the telephonic conversations held at airport with background aeroplane sound. Other examples are the conversation held in industrial environment with high noise machines, in railway stations etc. In all of these conditions, conversation (voice signal) get corrupted by background sound (noise) reducing its intelligibility. Thus, the need arises for an effective method that helps in reducing the background noise thereby increasing clarity of the conversation. Wavelet Transform (WT) is widely used for audio, image and video signal processing. This paper presents a wavelet-based approach to reduce the effect of background noise. This approach is compared with the standard, freely available softwares such as Audacity, Wavepad and Ocenaudio which are widely used for the purpose of noise reduction. The AWGN (Additive White Gaussian Noise) is assumed to be a model of background noise. Using extensive simulation study and analysis which is carried out using MATLAB
^®
2016a, WT based approach is found to perform better than the other standard softwares that are considered here. Correlation co-efficient is considered as a measure of performance improvement by the WT approach over the other software. The work has useful implications in various applications in which speech signal gets degraded by background noise.