Residual periodic signal evaluation algorithm after noise reduction

Abstract

The invention discloses a residual periodic signal evaluation algorithm after noise reduction, and the algorithm comprises the following steps: S1, setting a signal collection device to sample sound waves in equal-interval time, and carrying out the superposition simulation of the noise in a normal background database and the noise in a fault sound database with periodic fault noise; s2, performing high-dimensional feature extraction on the sampled sound wave, and extracting a residual signal of a construction signal; and S3, inputting the features extracted from the high-dimensional features into the similarity prediction network, and judging whether the periodic fault data is continuous or not. By combining a background noise library and a fault noise library, training a pre-processed signal periodicity estimation network, actually calculating the periodicity of a periodically sampled signal through a periodicity network, performing periodicity estimation through a transformer network, and aiming at the design of periodicity sampling equipment, the periodicity of a residual signal can be effectively evaluated, and the background noise problem of a specific scene can be solved.

Description

technical field [0001] The invention relates to a signal evaluation algorithm, in particular to a periodic signal evaluation algorithm remaining after noise reduction, and belongs to the technical field of signal processing. Background technique [0002] Noise is a kind of sound that causes irritability, or is too loud and endangers human health. From the perspective of environmental protection: any sound that hinders people's normal rest, study and work, or interferes with the sound that people want to listen to, All belong to noise, from the point of view of physics: Noise is the sound made when the sounding body vibrates irregularly. [0003] The general evaluation is to observe the periodicity through Fourier transform, or calculate the cross-correlation coefficient to observe the correlation, but because the background noise is different, the room is different, the signal distortion and distortion obtained after noise reduction will be different, so through the original...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a residual periodic signal evaluation algorithm after noise reduction, to solve the problem in the above background technology that the signal distortion and distortion obtained after noise reduction will be different due to different background noises and different rooms and venues. Therefore, through the principle Some correlation algorithm calculations cannot solve the problem of specific scenarios

Images