面向混合交通流的智能网联车鸣笛意图识别模型

为使混合交通流(Mixed Traffic Stream，MTS)下智能网联车(Intelligent Connected Vehicle，ICV)实现鸣笛意图(Horn’s Intention，HI)识别，更好地遵循常规车辆(Manual Vehicle， MV)的驾驶意图，提出ICV 对MV 鸣笛声的“ 感知(Perception) — 定位(Location) — 识别 (Recognition)”模型(简称HI-PLR)，采用深度卷积—循环神经网络(Deep Convolution Recurrent Neural Network, DCRNN)算法感知鸣笛车辆(Horning Vehicles, HV)的鸣笛声；采用到达时差 (Time Difference of Arrival, TDOA)算法定位HV；再基于运动时间窗(Motion Time Window, MTW)的支持向量机(Support Vector Machine, SVM)算法识别HI.实验结果表明，HI-PLR可使 ICV 对混流中车辆的鸣笛声感知准确率达90.4%，定位角度估计误差小于5°，HI 识别率达 82.5%，为ICV在MTS中的智能驾驶决策提供依据.

Identification Model for Horn's Intention of Intelligent Connected Vehicle under the Mixed Traffic Stream

With the aim that intelligent connected vehicles is able to identify horn's intention to follow the driving intention of the conventional vehicles better under the mixed traffic stream, perception-location-recognition model is proposed of ICV to the horn of conventional vehicles. Deep convolution recurrent neural network is used to percept the horn of the horning vehicles. time difference of arrival is exploited for the location of the HV. Support vector machine based on motion time window is applied to recognize the HI of the HV. The experimental results indicate that such a model enables the average accuracy rate of perception that the ICV conducts on the horn of the HV in the mixed traffic stream to amount to 90.4%, the error of positioning angle is within 5 degrees and the average recognition rate of HI is 82.5%, which provides the basis for the intelligent driving decision of the ICV in the mixed traffic stream.