基于子频带能量特征提取的汽车鸣笛声识别*

针对城市中汽车违法鸣笛声之间识别分类较难的问题，为了快速准确的识别鸣笛声并将不同种鸣笛声之间进行分类，在鸣笛声识别分类中提出了应用子频带能量提取鸣笛声的特征，并利用BP神经网络对提取的子频带能量特征值矩阵进行学习训练，且在神经网络学习过程中利用可变学习速度的方法，减小了神经网络的迭代次数。实验表明利用此种子频带能量特征提取法使鸣笛声与非鸣笛声的平均识别率达到了94.889%；使不同鸣笛声之间的分类正确率最大达到了93.75%，实现了不同鸣笛声之间的分类。

Recognition of automobile whistle sound based on sub-frequency band energy feature extraction

Aim at the classification of illegal car whistling in cities, in order to identify different kinds of car whistle quickly and accurately, the method of sub-frequency band energy feature extraction was applied in the classification and recognition of whistles. And the extracted sub-frequency band energy eigenvalue matrix was trained by BP neural network, and the number of iterations of the neural network in the process of learning was reduced by used the method of variable learning speed. The experiment shows that the average recognition rate of whistle and other sounds is 94.889%;and the classification accuracy rate of different car whistle sound is 93.75%, the classification among different whistles can realized by the method of sub-frequency band energy.