doi:10.16186/j.cnki.1673-9787.2020.6.18

Received:2019/12/26

Revised:2020/03/18

Published:2020/11/15

Research on environmental sounds recognition based on MFCC，MODGDF and SVM

REN Liyong1, HE Yongbin1, HE Xi1, YU Yongbin1, LIU Siyi2

1.School of Information and Software Engineering， University of Electronic Science and Technology of China ， Chengdu  610054 ， Sichuan， China;2.Sichuan Changjiang Vocational College，Chengdu  610106，Sichuan，China

Abstract:Environmental sounds recognition is an important research direction in the field of machine learning. It is used to help intelligent systems recognize environmental sounds from audio data. A new environmental sounds recognition method was proposed which combined Mel frequency cepstral coefficents （MFCC） and modified group delay function （MODGDF） as feature parameters， and then used multi-classification support vector machine （SVM） to classify the parameters to achieve the purpose of recognizing environmental sounds from audio data. The results showed that the experimental results of the DCASE 2018 datasets of the proposed method were better than the DCASE 2018 dataset baseline system recognition， and the recognition accuracy was 25.8% higher，respectively.

Key words:environmental sound recognition;Mel frequency cepstral coefficent;modified group delay function;support vector machine

基于MFCC_MODGDF和支持向量机的环境音识别研究_任立勇.pdf