Portable Classification System of Daily Activity and Exercise Load

本研究提出了一种新的算法，利用便携式测量系统对人体步行模式和日常活动进行分类和记录。将Inooka等人开发的人体步态分类方法编程到该系统中。该方法将人的步态分为平走、上楼梯和下楼梯。该系统对人的步态进行实时分类，并将估计的步态模式与人的心电图一起记录到动态心电图记录仪中。结果表明，该系统可以得到人的步行模式与心电图之间的关系。针对加速度和角速度积分计算的步行路径估计中存在数值误差，提出了人体步行路径的修正算法。本研究利用了误差椭圆内的估计动作和已知地标。如果一个路标与估计的动作有关系，则修改计算路径以通过该路标。例如，椅子作为地标的存在与坐下的动作有关。此外，我们还提出了一种将人类活动分为四种模式的方法：步行、休息、运动和躺着。这些活动是通过加速度的统计特性和加速度波形与已测得的参考波形的比较得出的。此外，连续活动是由行为转换的自动机决定的，它定义了人类行为的顺序。将自动机应用于运动轨迹和回溯，对人类活动的可靠估计是有效的。实验结果表明，使用该自动机可以很好地估计室内的人类活动。

This study proposes new algorithm to classify and record human walking pattern and daily activity by using a portable measurement system.Classification method of human gait developed by Inooka et al is programd in the system. The method classifies the human gait into walking flat, going up the stairs and going down the stairs. The system performs real-time classification of human gait and records the estimated walking patterns into the Holter ECG recorder with the ECG of human. It is suggested that the system yields the relation between human walking patterns and the ECG.Modification algorithm of human walking path is also proposed because numerical error is included in the estimated walking path calculated by the integration of acceleration and angular velocity. This study utilizes the estimated action and already-known landmarks within an error ellipse. If one landmark has a relation to the estimated action, then the calculated path is modified to pass the landmark. Existence of a chair as a landmark is related to action of sitting down, for example.Moreover, we proposed a method to classify the human activity into four kinds of patterns : walking, rest, movement and lying. These activities are derived by statistical characteristics of acceleration and comparison the acceleration wave form with reference wave form already measured. Furthermore, the continuous activity is determined by an automaton of behavior transition which defines sequence of human behavior. The application of the automaton in running its course and tracing back is effective to the reliable estimation of human activity. The experimental results show that the human activity in a room is favorably estimated by using the automaton.