Denoising Voip Calls Using Discrete Wavelet Transform

Abstract

Voice over Internet Protocol (VoIP) communication has become ubiquitous due to its cost-effectiveness and flexibility. However, the quality of VoIP calls can be significantly degraded by various types of noise present in the acoustic environment and introduced during transmission. This paper presents a novel approach to noise reduction in VoIP calls utilizing the Discrete Wavelet Transform (DWT). The proposed method effectively addresses both stationary and non-stationary noise sources, which are common challenges in real-world VoIP scenarios. By decomposing the noisy speech signal into different frequency sub-bands, DWT allows for targeted noise attenuation. We investigate various wavelet families and decomposition levels to identify the optimal parameters for noise suppression while preserving speech quality. The core of the algorithm involves applying a thresholding technique to the DWT coefficients, effectively distinguishing between speech and noise components. Coefficients below a defined threshold are either shrunk or set to zero, thereby reducing noise energy. Experimental results, evaluated using objective metrics such as Signal-to-Noise Ratio (SNR) improvement, demonstrate that the DWT-based noise reduction technique significantly enhances the clarity and intelligibility of VoIP calls, outperforming traditional noise reduction methods in diverse noisy environments. This research contributes to improving the overall user experience and reliability of VoIP communication.