Speech Enhancement Algorithms for Wind Noise Reduction

Speech is the most convenient communication mode. Speech signal, while transmission degrades due to various factors such as white noise addition, noise from the surrounding environment, etc. Wind noise gets added in mobile phones and corrupts speech signals leading to listener’s fatigue. Speech enhancement is a research area focusing on improving the quality and intelligibility of degraded speech signals. Wind noise, being non-stationary, is difficult to remove. This work focuses on reducing the effect of wind noise using various speech enhancement algorithms such as Multi-band Spectral Subtraction, Minimum Mean Square Error Short-Time Spectral Amplitude (MMSE-STSA), MMSE log STSA and Wiener filtering. The experiments are conducted considering wind noise alone and in combination with white noise. The performance of these speech enhancement algorithms on wind noise is evaluated using measures such as overall SNR, segmental SNR, Perceptual Evaluation of Speech Quality (PESQ), Itakura Saito Spectral distance and Short-Time Objective Intelligibility (STOI).