%0 Journal Article
%T 面向腕臂动作识别的加速度计数据分析<br>Elbow-wrist Activity Recognition via Accelerometer Data Analysis
%A 朱国康
%A 周涛
%J 机械科学与技术
%D 2018
%X 加速度传感器技术的不断发展和成熟为可穿戴设备的智能交互提供了新的途径。以可穿戴设备中的加速度计所捕捉的数据为支撑，对人的运动模式进行识别进而推断行为意图以实现智能化人机交互的方式，目前已受到人们的广泛关注。针对上述问题，本文基于循环神经网络的技术框架，提出一种高效的腕臂动作识别方法。该方法以单门控机制有效利用时序相关信息，对腕部可穿戴设备的三轴加速度计信号进行深度挖掘。实验结果表明，本文所提出的方法相较传统的机器学习方法而言性能更为可靠，能以更为简洁的网络结构取得与当前主流的循环神经网络算法同样的识别精度。<br>The present development of acceleration sensor technology provides a new way to the intelligent interaction of wearable devices. Nowadays, many researchers are interesting in identifying human activities through the data obtained by the accelerometers integrated in the wearable devices, and then inferring the behavior intentions for intelligent human-computer interaction. An efficient method of recognizing elbow-wrist activities is presented based on the recurrent neural network framework in this paper. The proposed method can deeply exploits the relevant sequential information embedded in the wearable tri-axis accelerometer data via a single-gated recurrent neural network. The experimental results show that the method proposed in this paper is more reliable than traditional machine learning methods, and is more concise in network structure to achieve an acceptable performance of elbow-wrist activity recognition compared with the current mainstream recurrent neural network algorithms
%K 可穿戴设备
%K 加速度计
%K 传感器
%K 循环神经网络
%K 动作识别&lt
%K br&gt
%K wearable device
%K accelerometer
%K sensor
%K recurrent neural network
%K activity recognition
%U http://journals.nwpu.edu.cn/jxkxyjs/CN/abstract/abstract7123.shtml