Environmental sound classification method, system and related device

Abstract

A method for classifying environmental sounds provided by the present application includes: collecting the environmental sounds in the target area to obtain audio files; performing spectrogram processing on the audio files to extract physical characteristic data corresponding to the environmental sounds; converting the physical characteristic data Input the preset hybrid classification prediction model, and output the classification result; wherein, the network structure of the preset hybrid classification prediction model is a combination of the network structure of the deep convolutional neural network model and the network structure of the Light GBM model. The network structure of the preset hybrid classification prediction model in this method is composed of the network structure of the deep convolutional neural network model and the network structure of the Light GBM model, that is, the preset hybrid classification prediction model integrates the deep convolutional neural network model With the advantages of the Light GBM model, it enhances the robustness and can improve the accuracy of environmental sound classification. The present application also provides an environmental sound classification system, equipment, and computer-readable storage medium, all of which have the above beneficial effects.

Description

technical field [0001] The present application relates to the field of environmental sound classification, in particular to an environmental sound classification method, system, device and computer-readable storage medium. Background technique [0002] With the rapid development of the Internet and information technology, people's living standards are improving day by day, and the quality of life and work requirements are also getting higher and higher. Audio, as a medium in people's daily life and business activities, deeply affects daily life. the act of living. Audio recognition is a cutting-edge research topic in the field of pattern recognition today. As a main research branch of audio recognition, Environmental Sound Classification (ESC) has recently attracted the attention of many experts and scholars and has become a hot topic. ESC is one of the most important technologies for machines to analyze their acoustic environment, and is widely used in surveillance, smart ...

AI Technical Summary

Problems solved by technology

Traditional machine learning-based methods often have unsatisfactory results in dealing with this problem, and the obtained models cannot effectively perform classification predictions; and the model structure based on shallow convolutional networks often trains models that still have to be trained. improve

The shallow convolutional neural network model has a single structure, and it is impossible to determine whether it has reached the optimal structure. Moreover, the model has a single structure and insufficient robustness, and the difference between multiple trainings is relatively large.

In short, the accuracy of classification results obtained by traditional methods for classifying environmental sounds is low

Images