基于手机传感器数据识别交通方式中最佳时间窗口选取研究

Selection of the Best Time Window in the Travel Mode Based on Mobile Phone Sensor Data

为探究利用手机传感器数据提取交通方式时,对原始数据处理时选取不同的时间窗口长度对出行方式识别的影响,设计多种交通方式的组合出行试验,利用手机传感器采集软件采集真实出行数据,同步记录出行日志。经过数据预处理与出行特征分析,得到不同时间窗口下的特征参数,并利用支持向量机算法对出行方式进行识别。结果表明,出行方式整体识别准确率随时间窗口长度呈先增大后减小的趋势,在30s~50s之间的准确性较高,35s时的识别准确性最高,达到83.63%。其中,步行、自行车、公交车、小汽车四类交通方式准确率可到达90.48%、84.21%、80.98%、78.85%。结论为利用手机传感器识别交通方式的准确性提供时间窗口选择上的依据。

In this paper,in order to explore the influence of the length of the time window selected by the original data processing on the travel mode identification when using the mobile phone sensor data to identify the travel mode.The combined travel mode of various modes of transportation was designed,and the real-time travel data was collected by the mobile phone sensor acquisition software,and the travel log was recorded synchronously.After data preprocessing and travel feature analysis,the feature parameters under different time windows are obtained,and the travel mode is identified by the support vector machine algorithm.The results show that the overall recognition accuracy of travel mode increases first and then decreases with the length of time window.Between 30s and 50s,the accuracy is higher,and the recognition accuracy is the highest at 35s,reaching 83.63%.Among them,the accuracy of walking,bicycle,bus,and car transportation can reach 90.48%,84.21%,80.98%,and 78.85%when the accuracy rate is 35s.The conclusion is to provide support for time window selection by using mobile phone sensors to identify the accuracy of the traffic mode.