Speech enhancement method based on Kullback-Leibler difference

Abstract

The invention discloses a speech enhancement method based on Kullback-Leibler difference and determines the optimal decomposition number of the method. The method is specifically characterized in thatfirstly, a noisy speech signal is processed into frames, each frame of signal is processed separately, and optimal atoms a1, a2,...and ak are selected through the KL difference principle; secondly, the best rational orthogonal base Bk is constructed according to the atoms and combined with a weight coefficient to obtain reconstruction signals fk<^>; thirdly, the reconstruction signals obtained through N times of decomposition are superposed to obtain a final denoised speech signal; and lastly, the RMSE attenuation difference is taken as a cost function to determine the optimal decomposition number. The method is advantaged in that the KL difference selection principle is utilized, the best atoms are adaptively selected, the base function is further constructed, the uncertainty caused by artificial parameter selection is greatly reduced, better noise reduction performance at the low SNR is achieved, secondly, the optimal decomposition number is determined according to the cost function, computational complexity is effectively reduced, and the method can be widely applied to fields such as speech noise reduction.

Description

technical field [0001] The invention relates to a speech enhancement method based on Kullback-Leibler (KL) difference, which is applied in the technical fields of wireless telephone communication, scene recording, military wiretapping and the like. Background technique [0002] The purpose of speech signal processing is to obtain certain speech characteristic parameters for efficient transmission or storage, or to achieve a certain purpose through a certain processing operation, such as artificial speech synthesis, identifying the speaker, and the content of the speech. Speech enhancement is an important aspect of speech signal processing. One of the main purposes of speech enhancement is to extract the pure original speech signal as much as possible from the speech signal mixed with noise. A pure speech signal is almost impossible, especially in the case of low signal-to-noise ratio. In this case, there are two main purposes of speech enhancement: one is to improve speech...

AI Technical Summary

Problems solved by technology

[0007] For the deficiencies in the prior art, the purpose of the present invention is to propose a speech enhancement method based on the Kullback-Leibler difference, and determine the optimal decomposition times of the method, which overcomes the poor noise reduction performance of the traditional method under the low signal-to-noise ratio , The shortcomings of too many parameters relying on empirical selection, which significantly improves the noise reduction performance in the case of low signal-to-noise ratio, adaptively selects parameters without relying on human factors, and selects the best decomposition times, reducing the amount of calculation

Images