Robust speech features extraction in convolutional noise environment

To resolve the mismatch between training features and testing features in convolutive noise environment,a robust speech features extraction algorithm based on Independent Component Analysis (ICA) is proposed.Noisy speech signals are firstly converted from time-domain to frequency-domain via short time Fourier transform,then a complex ICA algorithm is used to acquire short-time spectrum of speech signal from that of noisy speech signal,furthermore,Mel Frequency Cepstral Coefficients (MFCC) and its 1-order differential coefficients are computed in accordance with the separated speech signals frequency spectrum.Simulation and real environment experiments on different noisy Chinese digit recognition are carried out.The results show that the recognition ratio of the proposed algorithm obtains the relative increasing of 34.8%and 32.6%compared with conventional MFCC,which reveal that the speech features based on ICA have a good robust performance.