Engineering Commons^™

A Robust Noise Spectral Estimation Algorithm For Speech Enhancement In Voice Devices, Brett Joseph Chessher

LSU Master's Theses

In this thesis, a new robust noise spectral estimation algorithm is proposed for the purpose of single-microphone speech enhancement. This algorithm can generate the optimal noise spectral estimates in the Minimum Mean Square Error (MMSE) sense based on the speech statistics in the noisy environments. Compared to the well-adopted conventional noise spectral estimation method using the single-pole recursion, our proposed scheme is more reliable since the recursion coefficients are adaptable and optimal in the MMSE therein. We also propose a new accurate Resulting Signal-to-Noise Ratio (R-SNR) estimator as a quality measure to benchmark the existing noise spectral estimation techniques. This …