基于HHT的声频传感器车辆分类识别  

Audio sensor vehicle classification recognition based on HHT

作　　者：毕超[1] 冯玉田[1] 李园辉[1] 刁志蕙 

机构地区：[1]上海大学通信与信息工程学院,上海200444

出　　处：《电声技术》2016年第6期48-52,共5页Audio Engineering

基　　金：国家青年科学基金项目(61301027)

摘　　要：针对传统语音信号识别过程中出现的识别率较低或者是计算时间复杂度高的问题,提出了基于希尔伯特黄变换(HHT)的快速声频传感器车辆识别方法。该方法将HHT算法和语音信号特征提取中常用的梅尔倒谱系数(MFCC)相结合,形成一种新的特征提取方法。实验中,将这种方法分别与K-近邻算法(K-NN)、支持向量机算法(SVM)和稀疏表示分类算法(SRC)配合进行语音信号识别,结果表明,该特征提取方法与K-NN分类算法配合,在识别率和算法运行效率方面具有明显的优势。In view of the traditional voice signal recognition, it is found that the recognition rate is low in the process or the computational time complexity is very high.To solve this problem, a fast voice recognition method based on Hilbert Huang Transform （HHT） which use acoustic sensor networks is proposed.This method is mainly using the HHT algorithm combine with Mel Frequency Cepstrum Coefficient（MFCC） which is commonly used in voice signal feature extraction, to form a new feature extraction method.Finally this method is used to combine with traditional K-NN classification algorithm, SVM classification algorithm and SRC algorithm to complete voice signal recognition.The experimental results show that this feature extraction method with the K-NN classification algorithm has obvious advantage in the recognition rate and efficiency.

关 键 词：语音分类识别 希尔伯特黄变换(HHT) 特征提取 

分 类 号：TN912.3[电子电信—通信与信息系统]