A noise masking method with adaptive thresholds based on CASA

Abstract

In this paper, we propose a novel noise masking method based on Computational Auditory Scene Analysis by using an adaptive factor. Although it has succeeded in the field of speech separation and speech enhancement to some extent, the usage of fixed thresholds used for segregation and labeling heavily affects the processing performance. Focusing on this issue, the proposed method utilizes the Normalized Cross-Correlation Coefficients between the power spectra of noisy speech and pure noise to find an adaptive threshold, so that the pitch contour and Time-Frequency units can be obtained more accurately. Then, a revised algorithm is used to smooth the current binary mask value by checking the Time-Frequency units within adjacent frames and neighbor channels around the current Time-Frequency unit in order to remove the erroneous local masks. Two kinds of Signal to Noise Ratio test results show that the performance of the proposed method outperforms conventional spectral subtractive, Wiener Filtering and Computational Auditory Scene Analysis methods.