Computer Engineering and Applications ›› 2018, Vol. 54 ›› Issue (11): 128-132.DOI: 10.3778/j.issn.1002-8331.1701-0196

Previous Articles     Next Articles

Driving behavior recognition based on Sparse Filtering-Convolutional Neural Network

WANG Zhongmin, ZHANG Yao, HENG Xia   

  1. School of Computer Sience & Technology, Xi’an University of Posts & Telecommunications, Xi’an 710121, China
  • Online:2018-06-01 Published:2018-06-14

SF-CNN在驾驶行为识别中的应用研究

王忠民,张  瑶,衡  霞   

  1. 西安邮电大学 计算机学院,西安 710121

Abstract: In order to reduce the potential dangers and ensure the drivers safety, this paper proposes an unsupervised feature learning algorithm model for optimal feature distribution—Sparse Filter-Convolution Neural Network, which uses the built-in sensors of smartphone to assist the driver safety. This method utilizes the three-axis acceleration data acquired by smartphone in the running vehicle, and performs the normative joint constraint by sparse filtering to obtain the compact primary characteristic expression. As the input of the neural network, the nonlinear classification is used to identify the driving behavior. The experimental results show that this method has higher recognition rate and robustness to driving behavior, better than traditional neural network, and is very important to evaluate the efficiency of the assisted driving system.

Key words: sparse filter, convolutional neural network, pattern recognition, driving behavior

摘要: 为了减少不良驾驶行为的潜在危险,通过智能手机内置传感器对驾驶行为进行实时监测,辅助驾驶者安全驾驶,提出了一种优化特征分布的无监督特征学习算法模型——稀疏滤波-卷积神经网络模型(Sparse Filter-Convolutional Neural Network,SF-CNN)。该方法利用移动终端在车辆行驶中采集的三轴加速度数据,通过稀疏滤波进行范数联合约束,得到紧凑的初级特征表达,将该表达矩阵作为卷积神经网络首层的输入,进行非线性分类来识别驾驶行为。实验结果表明,稀疏滤波-神经网络的识别模型对驾驶行为具有更高的识别率和鲁棒性,优于传统神经网络模型,对辅助驾驶系统的效能评价有重要的理论意义。

关键词: 稀疏滤波, 卷积神经网络, 模式识别, 驾驶行为