交通事件小波分解与支持向量机声频识别方法

Traffic incident acoustic recognition method based on wavelet decomposition and support vector machine

分析了现有交通事件自动检测和识别方法,提出了应用小波分解与支持向量机相结合的交通事件声频识别方法。将车辆行驶的声音信号进行小波分解,以不同频段的重构信号能量作为特征向量,对由多个支持向量机构成的交通事件分类器进行训练,并对正常行驶、刹车和碰撞事件的声音信号进行识别。试验结果表明:利用车辆声音信号能够正确识别不同的交通事件,识别准确率达95%,识别方法可行。

The existing automatic detection and recognition methods of traffic incidents were analyzed,a recognition method with vehicle acoustic signals was proposed based on wavelet decomposition（WD） and support vector machine（SVM）.Vehicle acoustic signals were decomposed with WD,the powers in different frequencies were regarded as different incident eigenvectors,and the traffic incident classifier composed of multiple SVMs was trained.The acoustic signals of normal driving,braking and crash incidents were recognized.Test result shows that various traffic incidents can be recognized with vehicle acoustic signals,the recognition rate reaches 95%,so the proposed method is feasible.1 tab,3 figs,16 refs.