Noise Reduction from Speech Signal based on Wavelet Transform and Kullback-Leibler Divergence

A new method for speech enhancement basedon kullback-leibler (k-l) divergence has beenpresented in this paper. first, the algorithm performswavelet-packet transform to noisy speech anddecomposes it into sub-bands; then we apply a thresholdon coefficients in each sub-band to obtain enhancedspeech. to determine the threshold, first thedistributions of noisy speech, noise and clean speechcoefficients are calculated; then a symmetric k-ldivergence between the noisy speech and noisedistributions is calculated. finally a speech/noisedecision is made based on the calculated distance. weconducted some tests using timit database in order toassess the performance of the proposed method and tocompare it to previous speech enhancement methods.the algorithm is evaluated using the perceptualevaluation of speech quality measure (pesq) and theoutput snr. we obtain an improvement of up to 2.2dbon snr and 1.2 on pesq for the proposed method incomparison to the results of the previous wavelet basedmethods.