基于IMU的可穿戴式人体行为识别系统设计与实现

发布时间：2018-07-16 08:04

【摘要】：近年来,随着惯性测量单元(Inertial Measurement Unit,IMU)技术、无线体域网快速发展以及模式识别理论的成熟,基于可穿戴式技术的人体运动行为识别逐渐得到研究人员的重视,成为该领域的研究热点。相比于图像识别等识别方法,基于运动传感器行为识别显示出了功耗低、便携性好、成本低等优势,在医疗康复、人机交互、虚拟现实等领域都有了广泛应用。本文研究了基于可穿戴式多传感器运动信息融合的人体行为识别方法,实现了对日常人体运动模式的识别。研究的内容包括以及几个方面:(1)在现有惯性测量单元的基础上,设计了一种可穿戴式人体行为识别系统,该系统主要由微处理器、三轴加速度计、三轴陀螺仪、电源模块等组成,并通过低功耗蓝牙无线通信的方式实时、连续地向Android上位机传输加速度、角速度人体运动信息,并在Android平台上实现了人体运动信息的实时接收、动态显示和存储。(2)对比分析了基于加速度计求解姿态角和基于陀螺仪解算人体姿态角各自的优缺点,并针对传统人体姿态角解算算法精度低和稳定性差等缺点,提出了一种基于加速度计解算人体姿态角去校准四元数法求解姿态角的算法,将加速度和角速度数据进行有机融合,实时、准确地解算人体姿态角。(3)对人体运动信息进行时域分析和频域分析,以区分人体日常行为,并以人体运动数据采集实验为基础,将人体运动数据的时域特征特征、频域特征以及姿态角作为识别特征,提出了基于支持向量机多分类行为识别算法,并进行了人体运动模式识别。实验结果表明本系统能够实现人体日常行为的准确识别。
[Abstract]:In recent years, with the Inertial Measurement Unit (IMU) technology, the rapid development of the wireless body domain network and the maturity of the pattern recognition theory, the human body motion recognition based on wearable technology has gradually gained the attention of the researchers, and has become a hot topic in this field. The sensor behavior recognition shows the advantages of low power consumption, good portability and low cost. It has been widely used in medical rehabilitation, human-computer interaction, virtual reality and other fields. In this paper, the human behavior recognition method based on the motion information fusion of wearable multi-sensor is studied, and the research content of the daily human motion pattern is realized. Including and several aspects: (1) on the basis of the existing inertial measurement unit, a wearable human behavior recognition system is designed, which is mainly composed of microprocessors, three axis accelerometers, three axis gyroscopes, power modules and so on. Through low power Bluetooth wireless communication, the system is continuously transmitted to the Android host computer. Speed, angular velocity of human motion information, and on the Android platform, the real-time reception, dynamic display and storage of human motion information are realized. (2) the advantages and disadvantages of the attitude angle based on accelerometers and the gyroscope to calculate the attitude angle of the human body are compared and analyzed, and the accuracy of the algorithm is low and the stability is poor. In this way, an algorithm based on accelerometer to calculate the attitude angle of the human body attitude angle is proposed, which combines the acceleration and angular velocity data into an organic fusion, real-time and accurate calculation of the attitude angle of the human body. (3) the human body motion information is analyzed in time domain and frequency domain analysis to distinguish the human body's daily behavior, and the human body is transported by the human body. (four) the human body's daily behavior is distinguished from the human body and the human body is transported by the human body. On the basis of dynamic data acquisition experiment, the time domain feature, frequency domain feature and attitude angle of human motion data are used as recognition features, and a multi classification behavior recognition algorithm based on support vector machine is proposed, and human motion pattern recognition is carried out. The experimental results show that the system can realize the accurate recognition of human activities.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP212.9

【参考文献】