Audio classification system and method applied to abnormal sound detection

Abstract

The invention discloses an audio classification system and method applied to abnormal sound detection. Audio classification is an important module in an abnormal sound detection system. Existing audio classification methods like a Gaussian mixed model and a K-nearest neighbor are low in detection rate and cannot meet actual needs. Aiming at this situation, the audio classification system which uses a Probabilistic Latent Semantic Analysis (PLSA) model and a K-nearest neighbor classifier is put forward. In the system, firstly a signal feature vector is sent into the Probabilistic Latent Semantic Analysis model to be trained so as to obtain a sound subject term bag model, and then the K-nearest neighbor classifier (KNN) is utilized to conduct classification, and finally the system obtains a classified audio file. Compared with a traditional audio classification system which uses the K-nearest neighbor classifier, the audio classification system and method achieve a better classification effect.

Description

technical field [0001] The invention relates to an audio classification system, in particular to an audio classification system with improved classification accuracy in an abnormal sound detection system. Background technique [0002] Audio classification is an important module in the abnormal sound detection system, which can distinguish different audio signals according to audio features. At present, the main bottleneck affecting abnormal sound detection still exists. In terms of classification, it is mainly manifested in the traditional abnormal sound detection and classification methods such as GMM (Gaussian Mixture Model), HMM (Hidden Markov Model) and other methods. The detection rate is relatively low. The false alarm rate is high and cannot meet actual needs. Therefore, there is an urgent need for a new audio classification system to improve the accuracy of audio classification in the abnormal sound detection system, thereby improving the performance of the abnormal...

AI Technical Summary

Problems solved by technology

At present, the main bottleneck affecting abnormal sound detection still exists. In terms of classification, it is mainly manifested in the traditional abnormal sound detection and classification methods such as GMM (Gaussian Mixture Model), HMM (Hidden Markov Model) and other methods. The detection rate is relatively low. High false alarm rate, unable to meet actual needs

Images