Experimental Assessment of Spectral Subtraction and Kalman Filtering Algorithm on Electricity Generator Noise Reduction in Wireless Communication System

Abstract

Noise corrupts, slows down, and reduces the clarity or accuracy of communication. It prevents an undistorted signal or message from being transmitted over wireless communication systems. To reduce noise in wireless communication channels, methods like spectral subtraction and Kalman filter can be used. Spectral subtraction uses its subtractive ability to remove noise in a noisy speech signal, while Kalman filter provides estimates of some unknown noise variables given the measurements observed over time. Therefore, this paper proposes an experimental assessment of both Spectral subtraction (SS) and Kalman Filtering (KF) algorithm for an electricity generator noise-corrupted speech signal over a wireless communication system at the receiver’s end. The proposed assessment was carried out using noisy speech signals obtained from a conventional mobile phone and the evaluation was carried out using generator noisy speech signals recorded in a workshop where a generator is been used while, the experimental assessment was performed using MATLAB software. The noise attenuation techniques evaluation was analysed using Mean Square Error (MSE), Signal-to-Noise Ratio (SNR), Perceptual Evaluation of Speech Quality (PESQ), and Short-Time Objective Intelligibility (STOI). The analysis revealed that Kalman filter performed better than spectral subtraction in reducing generator noise in wireless communication systems.