基于融合特征的汽车鸣笛声识别方法

Recognition method of car whistle based on fusion feature

针对复杂路况和现实环境条件下,传统汽车鸣笛声识别方法的分辨力不足问题,提出了一种基于变分模态分解和融合特征的汽车鸣笛声识别方法。采用变分模态分解(VMD)将鸣笛声信号分解为多个固有模态信号(IMF),基于峭度准则筛选出主IMF分量并重构信号;提取重构信号的MFCC和LPCC特征参数。利用融合算法获得基于上述两种特征的融合特征参数,并将其作为BP神经网络模型的输入特征,实现汽车鸣笛声的准确识别。研究表明,相较于单一特征方法,基于融合特征方法可有效提取汽车鸣笛声特征,提升识别准确率。

Under complex road conditions and real environment conditions,the recognition methods of traditional car whistle usually fails.A car whistle recognition method based on variational mode decomposition and fusion features is proposed.Variational mode decomposition(VMD)is used to decompose the whistle signal into multiple intrinsic mode signals(IMF).Based on kurtosis criterion,the principal IMF components are selected and the signal is reconstructed.The MFCC and LPCC characteristic parameters of the reconstructed signal are extracted.The fusion feature parameters based on the above two features are obtained by using the fusion algorithm,which are used as the input features of BP neural network model to realize the accurate recognition of car whistle.The results show that compared with the single feature method,the fusion feature method can effectively extract the features of car whistle and improve the recognition accuracy.