A Novel Gait Pattern Recognition Method Based on LSTM-CNN for Lower Limb Exoskeleton

This paper describes a novel gait pattern recognition method based on Long Short-Term Memory(LSTM)and Convolutional Neural Network(CNN)for lower limb exoskeleton.The Inertial Measurement Unit(IMU)installed on the exoskeleton to collect motion information,which is used for LSTM-CNN input.This article considers five common gait patterns,including walking,going up stairs,going down stairs,sitting down,and standing up.In the LSTM-CNN model,the LSTM layer is used to process temporal sequences and the CNN layer is used to extract features.To optimize the deep neural network structure proposed in this paper,some hyperparameter selection experiments were carried out.In addition,to verify the superiority of the proposed recognition method,the method is compared with several common methods such as LSTM,CNN and SVM.The results show that the average recognition accuracy can reach 97.78%,which has a good recognition eff ect.Finally,according to the experimental results of gait pattern switching,the proposed method can identify the switching gait pattern in time,which shows that it has good real-time performance.

An Experimental Study on the Gait Patterns and Kinematics of Chinese Mitten Crabs

Despite the many studies on eight-legged animals and the importance of their mechanics of terrestrial locomotion, the mechanical energy of crabs in voluntary locomotion on uneven, unpredictable terrain surfaces has received little attention thus far. In this paper, motion video images of Chinese mitten crab （Eriocheir sinensis Milne-Edwards） locomotion on five types of terrains were recorded using a high-speed three-dimensional （3D） recording video system. The typical variables of locomotion such as gait patterns, duty factor, mechanical energy of the mass center, mass-specific write of the total mechanical power of the mass center, and percentage recovery, were analyzed. Results show that the Chinese mitten crab uses random gaits instead of the alternating tetrapod gait with the increasing terrain roughness. The duty factors of the rows of the leading legs are greater for all terrains than those of the rows of the trailing legs. On smooth terrain, the duty factors of the rows of the trailing legs are greater than that on rough terrains. Kinematic measurements and calculations reveal that similar to mammals, birds, and arthropods, the Chinese mitten crab uses two fundamental gaits to save mechanical energy： the inverted pendulum gait and the bouncing gait. The bouncing gait is the main pattern of mechanical energy conservation. The low probability of injury and energy expenditure due to adaptations to various terrains induce the Chinese mitten crab to modify the mass-specific rate of the total mechanical power of the mass center. The statistical results of percentage recovery also reveal that the Chinese mitten crab has lower energy recovery efficiency over rough terrains compared with smooth terrains.

Control strategy of central pattern generator gait movement under condition of attention selection

As a typical rhythmic movement, human being＇s rhythmic gait movement can be generated by a central pattern generator （CPG） located in a spinal cord by self- oscillation. Some kinds of gait movements are caused by gait frequency and amplitude variances. As an important property of human being＇s motion vision, the attention selection mechanism plays a vital part in the regulation of gait movement. In this paper, the CPG model is amended under the condition of attention selection on the theoretical basis of Matsuoka neural oscillators. Regulation of attention selection signal for the CPG model parameters and structure is studied, which consequentially causes the frequency and amplitude changes of gait movement output. Further, the control strategy of the CPG model gait movement under the condition of attention selection is discussed, showing that the attention selection model can regulate the output model of CPG gait movement in three different ways. The realization of regulation on the gait movement frequency and amplitude shows a variety of regulation on the CPG gait movement made by attention selection and enriches the controllability of CPG gait movement, which demonstrates potential influence in engineering applications.

特发性脊柱侧弯患者步行生物力学特征:三维运动捕捉的横断面分析

背景:脊柱侧弯的生物力学分析病例数量有限,仅限于对脊柱或下肢独立进行生物力学分析,缺乏对人体多维度整体的评估,难以反映躯干与下肢在日常活动中的运动关系,不利于综合指导临床治疗。目的:通过测量脊柱侧弯患者在步行活动中的脊柱运动学,研究脊柱不同节段及下肢运动学/动力学在步态活动中的关系,以全方位多层次评估脊柱侧弯患者与正常人群在生物力学的差异,为临床防治脊柱侧弯提供循证依据。方法:此研究为一项横断面研究,于2020年7月至2021年6月,在福州大学城、福建中医药大学附属康复医院纳入脊柱侧弯患者28例及同年龄段正常人28例。采用三维运动捕捉系统以100 Hz的采样频率捕捉步态的运动学数据,2块测力台(AMTI 400600,采样频率1500 Hz)嵌于长10 m(数据采集有效长度为4 m),宽2.4 m平地走道采集动力学数据。比较两组受试者执行平地步行时步态时空参数,运动学及动力学参数的差异。所有受试者在纳入后立即接受全脊柱X射线片测量比较两组差异。结果与结论:(1)脊柱侧弯患者肩胛带-躯干相对旋转活动度、胸廓-骨盆相对旋转活动度较正常组小(P<0.05),骨盆旋转活动度相较正常组大(P<0.05)。(2)脊柱侧弯患者髋、膝关节屈伸活动度,髋关节凹凸侧屈伸峰力矩、地反力凸侧内外向峰值较正常组小(P<0.05)。(3)脊柱侧弯患者的膝关节屈伸活动度非对称性指数、肩胛带-躯干相对旋转活动度非对称性指数、胸廓-骨盆相对旋转活动度非对称性指数较正常组大(P<0.05)。(4)上述结果证实,脊柱侧弯患者在平地行走时肩胛带-胸廓-骨盆之间僵硬的运动模式,髋膝关节屈伸活动度减少,髋关节双侧屈伸力矩峰值及地反力凸侧内外向峰值降低。这些特征可以作为康复评估及治疗方案制定的依据。

The HITCR-I:Evaluation on a free gait generation method for the hexapod robot on irregular terrain

The purpose of this paper is to present and evaluate a method of free gait generation for HITCRI,a hexapod walking robot.The HITCR-I is designed as a modularized structure of legs that is based upon a four-bar linkage mechanism and with force sensors in the tip of legs,distributed hardware structure and a modular software structure of the control system.Based on a set of local rules between adjacent legs,finite state machine(FSM) model is built to control the coordination of legs.An automatic smooth transition of gait pattern is achieved through deriving the mathematical relation between gait pattern and locomotion parameters.The disordered inter-leg phase sequence is adjusted to a regular state smoothly and quickly by the local rules based FSM,and the gait pattern can transform automatically adapting to irregular terrain.The experiment on HITCR-I has demonstrated that it can walk through irregular terrain reliably and expeditiously with the free gait controller designed in this paper.

Multi-level gait transition model for a new type of arthropod robot

To realize the continuous and variable gait transition for a new type of arthropod robot, a multi-level gait transition model is studied in this paper. The model is composed of central pattern generator （CPG） and saturation function. The CPG consists of four pairs of oscillators which can ex- hibit rhythmic activity when given stimulation signal S that lies in the range of saturation function. All oscillators receive the same S, but each pair of oscillators has different saturation functions. Multi- level gait transition can be realized when S changes regularly, as the oscillators start or stop oscilla- ting at different times. After computer simulation, the gait transition model is implemented in the ar- thropod robot. The experimental results show that ideal gait transition for the arthropod robot can be realized with the multi-level gait transition model.

Continuous and smooth gait transition in a quadruped robot based on CPG

To improve the smoothness of motion control in a quadruped robot, a continuous and smooth gait transition method based on central pattern generator （CPG） was presented to solve the unsmooth or failed problem which may result in phase-locked or sharp point with direct replacement of the gait matrix. Through improving conventional weight matrix, a CPG network and a MATLAB/ Simulink model were constructed based on the Hopf oscillator for gait generation and transition in the quadruped robot. A co-simulation was performed using ADAMS/MATLAB for the gait transition between walk and trot to verify the correctness and effectiveness of the proposed CPG gait generation and transition algorithms. Related methods and conclusions can technically support the motion control technology of the quadruped robot.

Exploring human rhythmic gait movement in the role of cerebral cortex signal

The rhythmic movement is a spontaneous behavior due to the central pattern generator （CPG）. At present, the CPG model only shows the spontaneous behavior, but does not refer to the instruction regulation role of the cerebral cortex. In this paper, a modified model based on the Matsuoka neural oscillator theory is presented to better show the regulation role of the cerebral cortex signal to the CPG neuronal network. The complex interaction between the input signal and other parameters in the CPG network is established, making all parameters of the CPG vary with the input signal. In this way, the effect of the input signal to the CPG network is enhanced so that the CPG network can express the self-regulation movement state instead of being limited to the spontaneous behavior, and thus the regulation role of the cerebral cortex signal can be reflected. Numerical simulation shows that the modified model can generate various movement forms with different modes, frequencies, and interchanges between them. It is revealed in theories that the cerebral cortex signal can regulate the mode and frequency of the gait in the ~ourse of the gait movement.

Balance control during gait initiation: State-of-the-art and research perspectives

It is well known that balance control is affected by aging, neurological and orthopedic conditions. Poor balance control during gait and postural maintenance are associated with disability, falls and increased mortality. Gait initiation-the transient period between the quiet standing posture and steady state walking-is a functional task that is classically used in the literature to investigate how the central nervous system(CNS) controls balance during a whole-body movement involving change in the base of support dimensions and center of mass progression. Understanding how the CNS in able-bodied subjects exerts this control during such a challenging task is a prerequisite to identifying motor disorders in populations with specific impairments of the postural system. It may also provide clinicians with objective measures to assess the efficiency of rehabilitation programs and better target interventions according to individual impairments. The present review thus proposes a state-of-the-art analysis on:(1) the balance control mechanisms in play during gait initiation in able bodied subjects and in the case of some frail populations; and(2) the biomechanical parameters used in the literature to quantify dynamic stability during gait initiation. Balance control mechanisms reviewed in this article included anticipatory postural adjustments, stance leg stiffness, foot placement, lateral ankle strategy, swing foot strike pattern and vertical center of mass braking. Based on this review, the following viewpoints were put forward:(1) dynamic stability during gait initiation may share a principle of homeostatic regulation similar to most physiological variables, where separate mechanisms need to be coordinated to ensure stabilization of vital variables, and consequently; and(2) rehabilitation interventions which focus on separate or isolated components of posture, balance, or gait may limit the effectiveness of current clinical practices.

基于CPG的六足机器人运动控制方法研究综述

以六足机器人为研究对象,分析了中枢模式发生器(CPG)的国内外发展现状,揭示其主要是以振荡器作为控制器进行信号输出。针对如何搭建CPG模型,以Hopf振荡器为例进行模型搭建和改进,搭建六足机器人实验平台验证CPG模型控制六足机器人步态运动是可行的。最后对基于CPG的六足机器人发展进行了展望。

基于SVM-RFE的动力下肢假肢运动模式识别

下肢运动模式识别是动力型假肢控制系统最重要的一部分,为了充分利用截肢者大腿残肢运动信息和地形坡度信息,通过安装在残肢接受腔和足面的IMU以及足底压力传感器构成识别信号源,从中提取时域特征值用于运动模式识别。通过构建层次SVM识别模型,并采用RFE算法对子分类器选择合适特征,最终完成对于平地行走、上楼梯、下楼梯、上斜坡、下斜坡五种步态模式的识别。实验结果表明,该方法的识别准确率达到了99.5%,因此基于SVM-RFE的动力下肢假肢运动模式识别方法是有效可行的。

基于CPG的并联腿四足机器人步态控制

并联腿结构的四足机器人具有承载能力强、刚度大、运动精度高的优点,但控制复杂。为更好地控制,提出了一种串联结构等效方法对并联腿进行了分析,结合中枢模式发生器(CPG)对等效的髋、膝关节进行控制,从而实现四足机器人不同步态,并利用ADAMS与MATLAB/SIMULINK软件搭建了四足机器人的虚拟样机与控制系统。仿真实验结果表明,串联等效分析方法能够应用于并联腿结构,CPG可实现四足机器人的Walk与Trot步态,验证了该方法的可行性与有效性。

基于Kimura神经元振荡器的六足仿生机器人步态控制研究

生物中枢模式发生器(CPG)对六足机器人的步态控制具有重大研究意义,为此提出一种基于Kimura神经元振荡器的六足机器人CPG步态控制方法。首先,以蜘蛛为仿生对象设计六足机器人机械结构并对其进行运动学解算;然后,基于Kimura神经元振荡器建立振荡器模型并对其参数整定;其次,针对机器人6条单腿的相位关系设计CPG网络模型;最后,通过计算机仿真工具和样机进行联合实验。结果表明,基于Kimura神经元振荡器生成的CPG网络模型输出信号幅值和相位差稳定,能够满足六足机器人步态控制需求,为六足机器人的步态控制提供一种可行性方案。

80例脑血管病偏瘫患者步行训练效果观察

80例偏瘫患者接受以理学治疗方法为主的步行训练,训练后总步行率达86.25％,较训练前增加了52.5％,步态优良率达68.42％。实践证明:通过提高对髋、膝、踝各关节功能积分改善其功能状态的训练,可以达到提高步行能力和步行质量的目的。

基于多源信息融合的膝上假肢步态识别方法

为了实现膝上假肢的有效控制,提出基于多源信息融合的步态识别方法。首先通过搭建人体下肢多源运动信息系统获取下肢表面肌电信号、腿部角度信号和足底压力信号。针对获取的信息,采用基于小波变换的空域相关滤波对肌电信号进行消噪并提取信号特征;选择大小腿、膝关节角度作为腿部角度信号特征;将足底压力信号通过阈值法提取有效特征。在特征提取基础上,分别利用BP神经网络和有限状态机对下肢运动信息进行步态识别,并将识别结果进行融合。实验验证了该方法在平地行走、上下楼梯模式下步态识别准确率均达到95%以上。

基于加速度的小波能量特征及样本熵组合的步态分类算法

针对传统的使用小波分解系数作为特征对走路、上楼、下楼进行分类时不能对具有相同强度加速度信号的步态样本进行分类的问题,提出了一种基于样本熵和小波能量相结合作为特征的分类算法。利用三轴加速度传感器采集走路、上楼、下楼3种步态下的上臂加速度信号,将信号进行小波分解,提取能量特征和样本熵特征,构建决策树分类器和贝叶斯分类器。决策树分类器和贝叶斯分类器的总体分类精度分别为75%和78.75%,使用样本熵与小波能量作为特征的分类精度比仅使用小波能量的分类精度提高了15.85%和19.17%。就步态分类精度而言,样本熵与小波能量相结合的方法优于仅使用传统小波能量方法。

基于过程神经网络的步态模式自动分类

为克服步态辨识中特征向量法存在的步态特征难于提取、计算量大和算法复杂等局限,提出一种基于过程神经网络的步态模式自动分类综合方案.为感知人体步态,在测试者下肢安装加速度传感器来采集步行过程中的时序运动学信息.采用巴特沃斯滤波处理并将其拟合为时变函数直接输入到过程神经网络,利用其对任意连续泛函的逼近能力来实现对不同步态模式下时序加速度信号的自动分类.同时,针对传统梯度下降法难于收敛和局部极小等问题,提出采用粒子群优化算法对网络的权函数和权系数进行学习.实验证明了该方案的正确性和有效性.

行走和跑步两种步态模式的实验分析

目的 :分析行走和跑步运动两种模式下人体步态的特征。方法及结果 :使用自行研制的三维运动检测分析系统对 5名健康男性在跑步机上进行了不同速率下行走和跑步 2种步态模式的比较实验。结论 :行走和跑步两种步态模式的参数具有明显差异 ,步速对步态参数的影响要小于步态模式的影响。

自然步态摆动期动力学协调模式的研究

建立了包含肌肉-肌腱动力特性和神经兴奋-肌肉收缩动力学方程在内的人体下肢生物力学模型,并针对自然步态摆动期的运动过程,采用最优控制方法进行了计算分析。计算结果与实验测量结果在运动学、肌电信号和肌肉活性参数等方面均具有很好的一致性;同时利用模型分析了肌肉协调运动模式,结果表明在自然步态摆动期运动过程中,神经系统将肌肉结成不同的协同肌群,并按照特定的时序结构控制不同肌群实现步行运动的加速、制动和姿态控制。

基于肌音信号的步态动作模式识别研究

通过采集腿部肌肉5个通道的肌音信号,利用3层决策树对跑步、上楼、下楼、走路、静止5种步态动作进行模式识别研究。在决策树的第1层和第2层,应用双阈值门限法识别静止和跑步两种步态模式,在第3层,提出基于步态信号的自适应不等长分割算法以及改进的模糊熵算法,利用线性分类器对走路、上楼、下楼进行分类识别。结果表明:双门限阈值法可有效地对静止和跑步进行识别,当采用改进的模糊熵特征时,对走路、上楼、下楼3种步态模式的分类准确率达到了94.87%;而当综合利用近似熵、样本熵和改进的模糊熵3种特征时,其分类准确率达到了98.76%。