基于运动传感器的老年人活动智能识别与应用开发

发布时间：2017-12-27 06:12

本文关键词：基于运动传感器的老年人活动智能识别与应用开发　出处：《重庆大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着普适计算技术的不断发展,人们对于舒适、安全、健康生活的需求也与日俱增。基于运动传感器技术的活动识别正在健康看护、人机交互、动作指导等领域发挥着重要的作用,同时活动识别也逐渐成为智能家居领域的一块重要的拼图。活动锻炼已成为老年人某些慢性疾病的辅助治疗手段之一,适当的运动有益于提升身体机能及减缓老化程度。通过对老年人的活动进行识别,从而计算老年人的活动量,分析其活动规律,有针对性的增加或减少某项活动,这对于老年人的健康看护是具有重要意义的。基于计算机视觉和运动传感器的活动识别是目前最为常见的两种方法。基于计算机视觉的方法存在易受干扰、监控范围局限、侵犯隐私等主要缺陷,而基于传感器的技术则具有抗干扰强、携带方便、数据获取自由、保护隐私等优点。本文提出了一种基于运动传感器和HMM的活动识别方法。针对老年人的活动类型及活动特点提取了标准差、能量等用以区分静态活动集合S中的活动,提取corr_VF、Amp、RAF(ratio forward)和RVF(ratio vertical forward)值用以区相似步态活动集D中的活动。在特征值提取后使用改进K均值算法生成HMM模型的观测值集合,然后定义了本文的活动识别模型。在经过Baum-Welch算法对HMM参数λ进行训练后使用Viterbi算法来进行老年人的活动识别。通过对比实验验证了本文方法能有效应用于老年人的活动识别,对于连续单一活动的平均准确率达到93.4%,尤其是对于相似活动其平均识别率达到了93.7%;对于随机组合序列活动的准确率达到91.1%。在实验过程中本文还对比了不同传感器佩戴部位对于活动识别精度的影响:在腰部、髋部等中枢部位佩戴传感器进行活动识别会取得高于四肢的平均活动识别精度。本文设计了一种针对于老年人的日常活动看护系统,该系统通过对活动识别记录进行统计分析,为老年人的看护人员(亲属、医护人员)提供图形化的活动统计、分析信息,并为亲属提供有价值的有助于老年人的日常生活照顾的专家知识。该系统对于提升老年人身体健康水平、生活质量具有重要意义。
[Abstract]:With the continuous development of pervasive computing technology, the demand for comfortable, safe and healthy life is increasing. Activity recognition based on motion sensor technology is playing an important role in health care, human-machine interaction, action guidance and other fields. Meanwhile, activity recognition has gradually become an important mosaic in the field of smart home. Exercise has become one of the auxiliary treatments for some chronic diseases of the elderly. Appropriate exercise is beneficial to improve the physical function and reduce the degree of aging. By identifying the activities of the elderly, we can calculate the activity of the elderly, analyze their activity rules, and increase or decrease a specific activity, which is significant for the elderly's health care. Activity recognition based on computer vision and motion sensor is the two most common method at present. Based on computer vision, there are main defects such as interference, limited monitoring range and privacy violation. Sensor based technology has the advantages of strong anti-interference, easy to carry, free data access, and privacy protection. In this paper, a motion recognition method based on motion sensor and HMM is proposed. According to the activity type and activity characteristics of the elderly, we extract standard deviation and energy to distinguish activities in static activity set S, extract corr_VF, Amp, RAF (ratio forward) and RVF (ratio vertical forward) values for activities in the similar gait activity set. After extracting the eigenvalues, the improved K mean algorithm is used to generate the set of observation values of the HMM model, and then the activity recognition model of this paper is defined. After training the HMM parameter by Baum-Welch algorithm, the Viterbi algorithm is used to identify the activities of the elderly. By comparing the experimental results indicate that this method can be used for identification of activities of the elderly, the average for the continuous single activity reach 93.4% accuracy rate, especially for the similar activities of the average recognition rate reached 93.7%; the accuracy rate for the random sequence of activities reached 91.1%. In the process of experiment, we also compared the influence of different sensors on the accuracy of activity recognition: wearing the sensor and identifying the activity in the waist, hip and other central parts would get the average activity recognition accuracy higher than the limbs. This paper introduces a design for the daily life of the elderly care system, this system through the statistical analysis of activity identification record for the care of elderly people (relatives, staff) provides a graphical statistical analysis, information activities, and provides expert knowledge of everyday life care for valuable help to the elderly the relatives. The system is of great significance to improve the health level and the quality of life of the elderly.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP212

【参考文献】