The reduction of noise in speech signals using wiener filter

This study focuses on the design and implementation of adaptive noise filtering for audio signals using the Wiener filter in MATLAB. Preprocessing steps including sampling, windowing, and normalization were applied to the noisy audio signals obtained from various real-world recordings. The Wiener filter algorithm was implemented to suppress noise and enhance signal quality. To evaluate its effectiveness, the mean squared error (MSE) and signal-to-noise ratio (SNR) metrics were used. The impact of parameter tuning, including window size and noise power estimation, was investigated. Initially, a simple noise model was assumed, and the Wiener filter was applied directly to the raw signals. The results showed significant noise reduction but highlighted the sensitivity of the filter's performance to accurate noise power estimation. To address this, an adaptive noise estimation technique was implemented, and the performance of the Wiener filter was reevaluated. The adaptive approach demonstrated improved noise suppression. Additionally, the Wiener filter's performance was compared to that of other noise reduction methods such as LMS filtering. The results showed that the Wiener filter consistently achieved better signal clarity and lower distortion. The study concludes that the Wiener filter is a robust and effective tool for audio signal enhancement.