基于梅尔频率倒谱系数和支持向量机的汽车鸣喇叭声识别

使用违法鸣笛辅助执法设备监测城市交通中汽车鸣喇叭事件的发生,可以有效地治理扰民的喇叭噪声,汽车鸣喇叭声的识别方法是其关键.为了准确高效地在交通噪声里识别出汽车鸣喇叭声,采用支持向量机(support vector machine,SVM)作为喇叭声和交通噪声的二分类器,针对汽车喇叭声的谐波特征分布特点,提取其梅尔频率倒谱系数(Mel frequency cepstrum coefficient,MFCC)作为特征向量,并分析MFCC的梅尔滤波器个数及特征维数对识别效果的影响.实验结果表明,通过增加MFCC特征中梅尔滤波器个数及特征维数可以改善识别效果,信噪比越低越明显.

Car Honking Recognition Based on Mel Cepstrum Coefficient and Support Vector Machine

Using Illegal Whistle Auxiliary Law Enforcement Equipment to monitor the occurrence of car honking events in urban traffic can effectively control the disturbing car honking. The recognition method of car honking is the key. In order to accurately and efficiently identify the car honking in road environmental noise, this paper uses support vector machine (SVM) as the binary classifiers of car honking and traffic noise. According to the harmonic feature distribution characteristics of car honk, the Mel frequency cepstrum coefficient (MFCC) is extracted as the eigenvector, and the influence of the number of Mel filters and feature dimensions on the recognition effect is analyzed. The results show that by increasing the number of Mel filters and feature dimensions of MFCC can improve the recognition effect, and such effect is more obvious for the lower signal-to-noise ratio (SNR).