Loudspeaker automatic classification method based on machine learning

Abstract

The invention provides a loudspeaker automatic classification method based on machine learning. The method comprises a training method of a neural network model and a test classification method. The training method of the neural network model comprises the following steps: testing a plurality of defective and non-defective loudspeakers through an electroacoustic test system, inputting obtained excitation signals and response signals as training samples into a neural network for training to obtain judgment specifications of each classification, and classifying the produced loudspeakers according to the classification principle; the test classification method comprises the following steps: testing the loudspeaker through the electroacoustic test system, inputting the obtained excitation signal and the response signal into the neural network model as characteristic samples, and comparing the characteristic samples with classification surfaces obtained in neural network model training to obtain a classification result. The classification method can accurately and objectively reflect the abnormal sound of the loudspeaker, automatically completes the automatic classification of abnormalsound detection, and replaces the manual listening and classification.

Description

[Technical field] [0001] The present invention relates to the technical field of speaker testing, in particular, a speaker automatic classification method based on machine learning. 【Background technique】 [0002] Frequency response is an important indicator that characterized the speaker performance, and it is easier to measure. The frequency response reflects the sensitivity of each frequency of the speaker the same excitation voltage. As a typical nonlinear system, the speaker is a thinner difference in the production process process leads to the difference in frequency response curves. Headphones and other products require at least one pair of speakers, so speaker producers need to divide the speakers with similar frequencies into the same group. The current mainstream scheme is based on the frequency response. [0003] During the quality inspection process on the production line, it is necessary to test all why the monomer is poor. The speaker monomer or speaker is likely to...

AI Technical Summary

Problems solved by technology

Enterprises need to cultivate "golden ears" for artificial listening. Artificial listening is subjective and relies on direct feedback from people. At the same time, the training period for "golden ears" is long and easy to fatigue

Images