基于神经网络模型的康复运动轨迹修正算法
|
摘要
在康复训练中记录患者的关节运动数据意义重大,关节点的运动轨迹可以由轨迹上一系列的特征点来描述。文章针对使用Kinect 2.0获得的下肢康复运动关节点数据中的误差、抖动等一系列问题,提出了一种加入检验判别的初始化机制和采用简单权值更新策略的快速神经网络模型来修正下肢康复运动的关节点数据。实验结果表明,该模型快速有效,且在运动轨迹随时间推移而改变时响应敏捷。
Records of patients' articulation data in rehabilitation training are significant.Trajectory joints can be described by a series of feature points on the track.In view of the errors,jitter and other problems in the joints data of lower limb rehabilitation exercise obtained by Kinect 2.0 sensor,a fast artificial neural network model using discriminating initialization mechanism and simple weights updating policy is presented with the purpose of correcting joints data.Experimental results show that the model is quick and efficient,and is able to respond agilely when the trajectory changes with time.
Records of patients' articulation data in rehabilitation training are significant.Trajectory joints can be described by a series of feature points on the track.In view of the errors,jitter and other problems in the joints data of lower limb rehabilitation exercise obtained by Kinect 2.0 sensor,a fast artificial neural network model using discriminating initialization mechanism and simple weights updating policy is presented with the purpose of correcting joints data.Experimental results show that the model is quick and efficient,and is able to respond agilely when the trajectory changes with time.
引文
[1]谢欲晓,白伟,张羽.下肢康复训练机器人的研究现状与趋势[J].中国医疗器械信息,2010,16(2):5-8.
[2]张毅,张烁,罗元,等.基于Kinect深度图像信息的手势轨迹识别及应用[J].计算机应用研究,2012,29(9):3547-3550.
[3]孟明,杨方波,佘青山,等.基于Kinect深度图像信息的人体运动检测[J].仪器仪表学报,2015,36(2):386-393.
[4]陈晓明.基于Kinect深度信息的实时三维重建和滤波算法研究[D].上海:上海交通大学,2013.
[5]陈理.Kinect深度图像增强算法研究[D].长沙:湖南大学,2013.
[6]BUCKLAND M,COLLINS M.AI techniques for game programming[M].Portland:Premier Press,2002:158-188.
[7]SHOTTON J,FITZGIBBON A,COOK M,et al.Real-time human pose recognition in parts from single depth images[J].Postgraduate Medical Journal,2011,56(1):1297-1304.
[8]蓝箭,孙娟,杨裕炳,等.基于Kinect骨骼追踪的控制方法的研究[J].仪表技术,2014(8):20-22.
[9]伍保华,王勇.一种下肢康复机器人机构及匀速训练方法的研究[J].合肥工业大学学报(自然科学版),2015,38(3):295-299.
[10]张丹,陈兴文,赵姝颖,等.基于Kinect骨骼预定义的体态识别算法[J].计算机应用,2014,34(12):3441-3445.
[11]黄建新.Kalman滤波的人体运动位置跟踪算法[J].华侨大学学报(自然科学版),2003,24(3):254-256.
[12]李义宝,张学勇,马建国,等.基于BP神经网络的改进算法研究[J].合肥工业大学学报(自然科学版),2005,28(6):668-671.
[2]张毅,张烁,罗元,等.基于Kinect深度图像信息的手势轨迹识别及应用[J].计算机应用研究,2012,29(9):3547-3550.
[3]孟明,杨方波,佘青山,等.基于Kinect深度图像信息的人体运动检测[J].仪器仪表学报,2015,36(2):386-393.
[4]陈晓明.基于Kinect深度信息的实时三维重建和滤波算法研究[D].上海:上海交通大学,2013.
[5]陈理.Kinect深度图像增强算法研究[D].长沙:湖南大学,2013.
[6]BUCKLAND M,COLLINS M.AI techniques for game programming[M].Portland:Premier Press,2002:158-188.
[7]SHOTTON J,FITZGIBBON A,COOK M,et al.Real-time human pose recognition in parts from single depth images[J].Postgraduate Medical Journal,2011,56(1):1297-1304.
[8]蓝箭,孙娟,杨裕炳,等.基于Kinect骨骼追踪的控制方法的研究[J].仪表技术,2014(8):20-22.
[9]伍保华,王勇.一种下肢康复机器人机构及匀速训练方法的研究[J].合肥工业大学学报(自然科学版),2015,38(3):295-299.
[10]张丹,陈兴文,赵姝颖,等.基于Kinect骨骼预定义的体态识别算法[J].计算机应用,2014,34(12):3441-3445.
[11]黄建新.Kalman滤波的人体运动位置跟踪算法[J].华侨大学学报(自然科学版),2003,24(3):254-256.
[12]李义宝,张学勇,马建国,等.基于BP神经网络的改进算法研究[J].合肥工业大学学报(自然科学版),2005,28(6):668-671.