论文标题
使用表示学习和利用基于小组的培训的无监督驾驶行为分析
Unsupervised Driving Behavior Analysis using Representation Learning and Exploiting Group-based Training
论文作者
论文摘要
驾驶行为监控在管理道路安全和降低交通事故风险方面起着至关重要的作用。驾驶行为受多种因素的影响，例如车辆特征，道路类型，交通类型，但最重要的是，个人的驾驶模式。当前的工作通过捕获驾驶模式的变化来执行强大的驾驶模式分析。它通过使用多层SEQ-2-SEQ自动编码器学习压缩时间序列（自动编码的紧凑型序列）来形成一致的组，并利用层次集群以及建议选择最佳距离度量。一致的小组有助于确定数据集中捕获的个体驱动模式的变化。这些组都是针对火车和隐藏的测试数据生成的。使用火车数据形成的一致组，用于培训分类器的多个实例。获得的最佳距离测量选择用于选择最佳的火车测试对一致的组。考虑到IMU传感器（加速度计和陀螺仪）捕获的信号以对驾驶行为进行分类，我们已经对公开可用的UAH-DRIVESET数据集进行了实验。我们观察到提出的方法，明显优于基准性能。
Driving behavior monitoring plays a crucial role in managing road safety and decreasing the risk of traffic accidents. Driving behavior is affected by multiple factors like vehicle characteristics, types of roads, traffic, but, most importantly, the pattern of driving of individuals. Current work performs a robust driving pattern analysis by capturing variations in driving patterns. It forms consistent groups by learning compressed representation of time series (Auto Encoded Compact Sequence) using a multi-layer seq-2-seq autoencoder and exploiting hierarchical clustering along with recommending the choice of best distance measure. Consistent groups aid in identifying variations in driving patterns of individuals captured in the dataset. These groups are generated for both train and hidden test data. The consistent groups formed using train data, are exploited for training multiple instances of the classifier. Obtained choice of best distance measure is used to select the best train-test pair of consistent groups. We have experimented on the publicly available UAH-DriveSet dataset considering the signals captured from IMU sensors (accelerometer and gyroscope) for classifying driving behavior. We observe proposed method, significantly outperforms the benchmark performance.
