Abstract: This paper proposes a motion pattern recognition algorithm which only uses acceleration transducer on phone. The algorithm executes frequency domain filtering of acceleration transducer signal collected and then adopts improved DBSCAN algorithm when clustering. The experiment results show that the algorithm achieves higher accuracy and availability.

Key words: acceleration transducer, movement pattern, filtering, clustering

摘要： 提出一种用于识别人体运动行为模式的算法，该算法仅需从智能手机加速传感器获取信号数据，对信号进行频域滤波，采用改进的DBSCAN算法进行聚类和识别出运动模式。实验结果表明该算法具有较高的准确率和实用性。

关键词: 手机加速传感器, 运动模式, 滤波, 聚类