基于惯性传感器的坐姿检测系统设计与实现

发布时间：2018-05-04 07:49

本文选题：坐姿识别 + 卡尔曼滤波　；参考：《哈尔滨理工大学》2017年硕士论文

【摘要】：坐姿是现代职业人群工作中最常见的姿势,在日常工作中保持健康坐姿对于老年人群预防疾病有着重要的意义。因此,正确的坐姿成为提高健康质量的重要指标。近年来,随着微机械电子系统技术的发展,特别是加速度和角速度传感器的出现,对坐姿检测的研究起到很大的推进作用。单一传感器的姿态检测,容易因惯性器件的漂移和累计误差,导致测量精度的下降。因此,通过多个传感器进行姿态检测,并借助数据融合算法提高姿态检测精度成为目前研究的热点。与基于计算机视觉技术的人体姿态识别方法相比,基于传感器的人体姿态识别和监测技术具有不泄露使用者的隐私、携带方便简洁、检测准确等诸多优点。本文在国内外研究发展的现状的基础上,进一步加深对坐姿检测的研究。在本文中利用MPU6050传感器、单片机、选择器和电源等硬件组成坐姿检测系统。通过传感器采集到的三轴加速度和三轴角速度数据,将数据通过I2C总线协议传到单片机,单片机再将数据上传到电脑中并储存。通过MATLAB软件先对采集到的原始数据进行处理,处理的方法是利用卡尔曼滤波方法进行滤波处理。再对处理后的数据利用四元数算法进行角度合成,之后利用得到的数据对其进行特征项的提取,利用SVM分类器对处理过的数据进行识别,最后达到对不同坐姿姿态的识别。本系统实验结果表明,本系统可以识别出不同身高不同体重的人的正坐、前倾、后倾、左倾、右倾以及左翘腿和右翘腿的坐姿姿态,识别率可以达到90%以上。
[Abstract]:Sitting posture is the most common posture in the work of modern professional population. It is of great significance to keep a healthy sitting posture in daily work for the elderly to prevent diseases. Therefore, the correct sitting position is an important indicator to improve health quality. In recent years, with the development of MEMS technology, especially the emergence of acceleration and angular velocity sensors, the research of sitting posture detection plays a very important role. The attitude detection of a single sensor is easy to be caused by the drift and accumulative error of the inertial device, which leads to the decrease of the measurement accuracy. Therefore, attitude detection based on multiple sensors and data fusion algorithm to improve the accuracy of attitude detection has become a hot topic. Compared with the human body attitude recognition method based on computer vision technology, the sensor based human posture recognition and monitoring technology has many advantages, such as not revealing users' privacy, convenient and simple to carry, accurate detection and so on. Based on the current situation of domestic and international research, this paper further deepens the research of sitting posture detection. In this paper, the MPU6050 sensor, single-chip microcomputer, selector and power supply hardware composed of sitting position detection system. The data of three-axis acceleration and triaxial angular velocity collected by the sensor are transmitted to the single-chip computer through the I2C bus protocol, and the data are uploaded to the computer and stored by the single-chip microcomputer. The original data is processed by MATLAB software, and the Kalman filter is used to process the original data. Then the processed data are angled synthesized by quaternion algorithm, and then the feature items are extracted by the obtained data, and the processed data are identified by SVM classifier, and finally the attitude recognition of different sitting posture is achieved. The experimental results of the system show that the system can recognize the sitting posture of people with different height and weight, including the positive, forward, backward, left, right, left and right legs, and the recognition rate can reach more than 90%.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP212

【参考文献】