A modified power-level-difference-based noise reduction for dual-microphone headsets

This paper studies the performance of the power-level-difference(PLD)-based two-channel noise reduction for dual-microphone headsets by using Monte Carlo simulation method. The analysis results indicate that both the accuracy of the noise power spectral density (NPSD) estimation and the smoothing factor have significant impacts on the amount of noise reduction. Based on these analysis results, a modified PLD-based two-channel noise reduction algorithm is proposed, where two schemes are introduced to improve the performance of the traditional PLD-based algorithm. In the first scheme, a non-stationary noise estimator is proposed to improve the tracking capability of the NPSD estimator. In the second scheme, an adaptive updating smoothing factor (AUSF) is used to reduce the variance of the spectral estimation and increase the amount of noise reduction. Experimental results with recorded data show better performance of the proposed algorithm.