Signal processing and interpretation of physical activity time series

本项目旨在评估精神和心身疾病患者和健康青少年的生态测量心理症状与身体活动数据之间的关系，并开发提取与症状恶化相关的改变的身体活动模式的方法。(1)慢性疲劳综合征（CFS）患者的研究我们提出了一种基于小波变换表征身体活动时间序列的新方法，通过变换的负模极大值来评价活动中断时的时间相关性。我们发现，在35分钟内，CFS患者的局部相关特性明显小于对照组，这表明CFS患者有更多的身体活动突然中断。(2)对初中生的研究结果：初中生的主观情绪和疲劳与每次测量前30 min的活动数据计算的统计指标存在显著相关。在包含活动数据统计的多元回归模型中估计主观疲劳，平均和非趋势偏度及其相互作用项是高度显著的疲劳估计量。与活动突然下降或中断有关的非趋势偏度表现出最高的显著性。(3)重度抑郁症患者的研究我们发现，在超过20年的时间里，在对照组和重度抑郁症患者中，休息时间持续小于总体活动记录平均值的时间服从幂律累积分布（从2分钟到200分钟）。重度抑郁症患者的幂律指数明显小于健康个体，这意味着患者的休息时间更长，发作更频繁。这些结果(1)-(3)表明，在体力活动中断时，可以通过统计参数来量化和预测精神和心身症状。少

The aim of this project was to evaluate the relationship between ecologically measured psychological symptoms and physical activity data in patients with psychiatric and psychosomatic disorders and healthy adolescents, and to develop methods to extract the altered physical activity patterns related to the symptom worsening.(1)Research on patients with chronic fatigue syndrome (CFS)We developed a new method based on the wavelet transform to characterize the physical activity time series, which can evaluate the temporal correlation at the interruption of activities by the negative modulus maxima of the transform. We found that up to 35 min the local correlation property was significantly smaller in CFS patients than in controls, suggesting that CFS patients have more abrupt interruptions of physical activity.(2)Research on junior high school studentsWe found that subjective moods and fatigue in these subjects had significant correlations with statistical indices calculated from 30 min ac … More tivity data before each measurement. In a multiple regression model incorporating statistics of activity data to estimate subjective fatigue, the mean and detrended skewness and their interaction term were highly significant as fatigue estimators. The detrended skewness which is related to abrupt drop or interruption of activities showed the highest significance.(3)Research on patients with major depressionWe found that durations of resting periods having continuously smaller counts than the overall recorded average of activity obey a power-law cumulative distribution over more than two decades (from 2 min to 200 min) in both control subjects and major depression patients. The power-law exponent of major depression patients was significantly smaller than that of healthy individuals, implying more frequent episodes of longer resting periods in the patients.These results (1)-(3) suggest that psychiatric and psychosomatic symptoms can be quantified and predicted by the statistical parameters at the interruption of physical activity. Less