基于车载手机数据深度特征的驾驶行为识别

针对驾驶行为识别问题,利用智能手机传感器采集相应车辆的加速度、角速度信息,并用手机角度信息对原始数据进行矫正处理。传统的驾驶行为识别方法须事先对原始数据单元人为进行特征提取。为改善繁琐的人工特征提取方法,提出一种驾驶行为识别领域基于改进的卷积神经网络的特征提取方法。原始数据经过组合后,作为卷积神经网络的输入。通过改变卷积神经网络的损失函数,提高类内样本特征的相似度,再将提取的特征作为核极限学习机的输入。实验结果表明,该方法可有效识别车辆的静止、急加速、急减速、正常行驶、左转弯、右转弯等驾驶行为。

DRIVING BEHAVIOR RECOGNITION BASED ON DATA DEPTH FEATURES OF VEHICLE MOBILE PHONE

Aiming at the problem of driving behavior recognition,this paper used smart phone sensor to collect acceleration and angular velocity information of the vehicle,and used mobile phone angle information to correct the original data.The traditional driving behavior recognition method must extract features from original data unit artificially in advance.To improve the tedious manual feature extraction method,we proposed a feature extraction method based on the improved convolution neural network in the field of driving behavior recognition.After the original data was combined,it was used as the input of convolution neural network.By changing the loss function of the convolution neural network,the similarity of the sample features in the class was improved,and the extracted features were used as the input of the kernel extreme learning machine.The experimental results show that the method can effectively identify the driving behavior of vehicles such as static,rapid acceleration,rapid deceleration,normal driving,left-turn,right-turn,etc.