Adaptive Filtering Methods and Adaptive Filters

Abstract

The invention discloses a self-adaptive filtering method and a self-adaptive filter. An embodiment of the self-adaptive filtering method comprises the steps of: receiving an input speech signal by using the self-adaptive filter; filtering the speech signal; constructing a first filter coefficient updating influence item according to a noise component, and constructing a second filter coefficient updating influence item according to a speech component; and updating filter coefficients of the self-adaptive filter based on the first filter coefficient updating influence item and the second filter coefficient updating influence item. The self-adaptive filtering method and the self-adaptive filter achieves the effect of regarding a speech distortion degree as a constraint condition of filter coefficient iterative updating, so that the self-adaptive filter gives consideration to the speech distortion degree while operating normally, and the magnitude of speech signal distortion is limited, thereby reducing the speech distortion degree while ensuring speech signal enhancement, and further eliminating the influence on speech recognition effect due to excessive distortion of speech signals in the speech recognition process.

Description

technical field [0001] The present application relates to the field of computers, in particular to the field of speech signal processing, and in particular to an adaptive filtering method and an adaptive filter. Background technique [0002] Adaptive filters are widely used in speech recognition systems. For example, in a far-field speech recognition system, the adaptive filter can be used to eliminate noise, reverberation and other interference in the far-field speech signal, improve the speech quality, and improve the recognition performance of the speech recognition system. The adaptive filter updates the filter coefficients in real time according to specific criteria, automatically and continuously works on a given input signal, and obtains the desired output. Currently, an adaptive filter generally adopts a method of updating filter coefficients: taking the entire error signal as a parameter to update the filter coefficients. [0003] However, when the above-mentioned...

AI Technical Summary

Problems solved by technology

[0003] However, when the above-mentioned adaptive filter is used for filtering, while eliminating the interference component in the speech signal, it also damages the speech signal itself, resulting in distortion of the effective speech signal

As a result of speech distortion, the speech signal does not match the speech recognition model, resulting in a drop in speech recognition rate

Images