Abstract: This paper designs a wearable posture recognition system based on kernel-based Support Vector Machines(SVM). With the characteristic parameters extraction of tilt sensors embedded in garments, it completes posture classification using two kinds of classification algorithms. Experiment which is subjected to five testers shows that using kernel-based SVM can give the best performance, and the designed system can make a good recognition rate when classifying everyday posture, which has a considerable value in the remote monitoring field.

Key words: kernel function, Support Vector Machines(SVM), wearable, posture recognition, remote monitoring

摘要： 设计一种基于核函数支持向量机(SVM)的穿戴式姿态识别系统。采集嵌入用户服装中的倾角传感器的数据，提取相应的特征参数，利用2种分类算法对样本进行姿态分类评估。测试实验结果表明，核函数SVM算法对日常姿态的分类效果较好，姿态识别系统对用户日常的多种姿态识别率较高。

关键词: 核函数, 支持向量机, 穿戴式, 姿态识别, 远程监护

CLC Number: 

• TP311.52