基于心理声学及支持向量机的扬声器异常音检测算法  

Speaker Abnormal Sound Detection Algorithm Based on Psychoacoustic Model and Support Vector Machine

作　　者：郭庆[1] 何劼恺 苏海涛[1] 王兴建 GUO Qing;HE Jiekai;SU Haitao;WANG Xingjian(School of Electronic Engineering and Automation,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《东华大学学报（自然科学版）》2020年第2期275-281,共7页Journal of Donghua University(Natural Science)

基　　金：广西自然科学基金资助项目(2016GXNSFBA380117);桂林市科学研究与技术开发资助项目(2016010404-3);厦门大学水声通信与海洋信息技术教育部重点实验室开放课题资助项目(201601);广西科技基地和人才专项资助项目(桂科AD19110026)。

摘　　要：目前扬声器异常音检测中,主要使用人工听音和工程师依据经验设置门限法,受主观因素影响大,且不能实现扬声器异常音的分类。为此,提出了一种新的扬声器质量评价方法,即基于心理声学模型和粒子群优化的支持向量机扬声器异常音检测方法。提取并标记扬声器声音响应信号,将其输入心理声学模型,得出心理声学能量均值并输入支持向量机;利用粒子群算法进行调优,最终得到具有最优参数的支持向量机。经试验验证,该模型的检测准确率达到98%。与音色特征法相比,其检测准确率得到较大的提高并实现了异常音分类。The artificial listening detection method or critical threshold method which value is set by the engineers with their experience are widely used in the speaker abnormal sound detection.They are severely influenced by subjective factors and cannot classify faulty speakers.A new speaker quality evaluation(SQE)based on psychoacoustic model and support vector machine(SVM)by particle swarm optimization(PSO)was proposed in order to overcome the above disadvantage.The sound response signal of a speaker was extracted and marked,then input into the psychoacoustic model.The average energy for the output of the psychoacoustic model was input to SVM.POS algorithm was used to obtain SVM with optimal parameters.Experiments show that detection accuracy of SQE reaches 98%.Compared with the timbre characteristics method,SQE improves the detection accuracy greatly and can classify faulty speakers.

关 键 词：扬声器 异常音检测 心理声学模型 支持向量机 

分 类 号：TN912[电子电信—通信与信息系统]