基于鲁棒纹理特征的环境声音事件检测方法  

Environment Sound Event Detection Method Based on Robust Texture Features

作　　者：吴婷 刘琼 郭慧茹[2] WU Ting;LIU Qiong;GUO Huiru(College of Mathematics and Physics,Shanghai Electric Power University,Shanghai 201399,China;School of Merchant Shipping,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海电力大学数理学院,上海201399 [2]上海海事大学商船学院,上海201306

出　　处：《电子器件》2024年第2期530-535,共6页Chinese Journal of Electron Devices

基　　金：国家自然科学基金项目(11801356)。

摘　　要：针对各种类别的环境声音事件检测问题,提出了基于鲁棒纹理特征的环境声音事件检测方法。首先,将原始的声音样本转换为类伽马声谱图;然后将类伽马声谱图通过剪切波变换提取图像的纹理特征;又采用中心化二值模式(CBP)算法进行编码;针对特征维度过高问题,先利用随机森林算法后结合主成分分析(PCA)算法,提出了RF-PCA降维方法;最后使用支持向量机(SVM)对不同环境的声音进行分类。在公开数据集ESC-10上的仿真实验结果表明,利用所提出的基于鲁棒纹理特征的环境声音事件检测方法所提取的特征对声音分类可达到93.00%的分类效果。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.

关 键 词：环境声音分类 类伽马声谱图 SHEARLET变换 CBP算法 RF-PCA 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]