Environment Sound Event Detection Method Based on Robust Texture Features
Aiming at various kinds of environmental sound event detection problems,an enviromental sound event detection method based on robust texture feature is proposed.Firstly,the original sound samples are converted into gammatone-like spectrogram.Then,the gammatone-like spectrogram is transformed by using Shearlet transform.The centralized binary pattern(CBP)algorithm is used for cod-ing.Aiming at the problem of too high feature dimension,random forest algorithm is first used and then principal component analysis(PCA)algorithm is combined,and RF-PCA dimension reduction method is proposed.Finally,support vector machine(SVM)is used to classify sounds in different environments.Results of simulation experiments on the public dataset of ESC-10 show that the features ex-tracted by using the proposed environmental sound event detection method based on robust texture feature can achieve 93.00%classifi-cation effect on sound classification.
国家科技期刊平台
NSTL
万方数据
