Feature Compensation Combining SNR-Dependent Feature Reconstruction and Class Histogram Equalization

Abstract

In this letter we propose a new histogram equalization technique for feature compensation in speech recognition wider noisy environments. The proposed approach combines a signal-to-noise-ratio-dependent feature reconstruction method and the class histogram equalization technique to effectively reduce the acoustic mismatch present in noisy speech features. Experimental results from the Aurora 2 task confirm the superiority of the proposed approach for acoustic compensation.