Robust Speech Recognition Based on Mixed Histogram Transform and Asymmetric Noise Suppression

This paper proposes a new feature extraction algorithm which is robust against noise using histogram compensation and asymmetric filter. Temporal masking would be provided to improve ASR systems specifically in matched and multistyle training conditions. Nonlinear filtering and temporal masking are used in this algorithm. By matching the power histograms of the input in each frequency band to those obtained over clean training data, and then mixing together the processed and unprocessed spectra can be increased appropriately speech recognition accuracy. Obtaining results show that recognition accuracy in compare with MFCC, PLP and PNCC has been improved in various training conditions.

Keywords:

en