GA-BP神经网络在下肢运动步态识别中的应用研究

为了提高下肢表面肌电信号步态识别的准确性,提出了一种基于遗传算法（GA）优化的BP神经网络分类器设计方法。首先,对采集的下肢表面肌电信号进行小波滤波及特征值提取,其次,构造基于GA优化的BP神经网络分类器,然后,以提取的表面肌电信号特征作为输入对分类器进行训练,最后利用训练好的分类器进行测试。实验结果表明,基于GA优化的BP神经网络分类器能成功识别下肢正常行走的五个步态,平均识别率达到98%以上,效果明显优于BP神经网络分类器的识别效果。     

基于MPSO-BP神经网络方法的人体步态识别

为提高人体下肢步态相位识别准确率以实现外骨骼机器人控制,采用一种改进的粒子群优化MPSO-BP神经网络方法识别不同运动模式下的人体步态相位。通过自适应调整学习因子构造MPSO-BP神经网络分类器,以多种传感信息组成的特征向量样本集训练神经网络分类器,用于识别人体下肢在平地行走、上楼梯和起坐三种典型运动模式下的步态相位。实验结果表明,MPSO-BP神经网络分类器能有效识别三种不同运动模式的步态相位,识别准确率均达到96%以上,识别性能优于传统的BP神经网络模型和粒子群优化神经网络模型。     

基于径向基函数神经网络的红外步态识别

为提高红外步态识别的效果,提出一种基于径向基函数神经网络的多分类器融合算法。对红外步态序列,分别应用基于轮廓线傅立叶描述子特征的模糊分类器和基于下肢关节角度特征的贝叶斯分类器进行识别,再利用径向基函数神经网络的学习和分类功能,对获得的输出信息进行度量层的融合和再识别。仿真实验结果表明,该算法获得更加精确的分类效果。     

利用多源运动信息的下肢假肢多模式多步态识别研究

运动状态识别对智能下肢假肢的控制非常关键,本文利用下肢表面肌电信号、腿部角度和足底压力信号在运动模式和步态分析中的优势和特点,对下肢假肢的多模式多步态识别进行研究.通过建立下肢运动信息系统,获取下肢多源运动信息.先提取下肢肌电信号的小波包能量作为特征,建立多个HMM对下肢假肢的运动模式进行识别;再根据大小腿和膝关节的角...     

结合双流网络和金字塔映射的步态识别

目前,基于深度学习的步态识别方法虽然取得了一定的进展,但数据采集和步态外观的变化仍然是实现精确步态识别所面临的挑战。为了提高网络对时空步态信息的捕捉能力,提出了一种基于步态轮廓流和步态特征差分流的双流网络结构。步态轮廓流以步态轮廓图作为输入,用来提取步态序列中包含的空间步态信息;步态特征差分流则是以步态特征差分图作为输入,用来捕获相邻步态图之间的动态信息。同时,为了充分利用步态序列中的全局和局部信息,提出了多尺度金字塔映射（multi-scale pyramid mapping,MPM）模块,并插入到各单流网络中以增强网络对全局和局部步态信息的提取能力。所提方法在步态数据集CASIA-B和OU-MVLP上的平均识别精度分别达到了87.0%和85.5%,这表明双流网络架构和MPM模块可以有效地捕获步态序列中的时空步态信息。     

基于CNN和DLTL的步态虚拟样本生成方法

针对步态识别在反恐、安防领域亟待解决的小样本问题,提出了一种基于深度卷积神经网络（convolutional and neural network,CNN）和DLTL（dual learning and transfer learning）的步态虚拟样本生成方法。首先用基于VGG19的深度卷积神经网络模型低层响应提取步态风格特征图,然后利用基于对抗网络的对偶学习（dual learning,DL）对风格特征图进行风格训练,得到风格特征模型;其次利用VGG19模型的高层响应提取步态内容特征图,然后让步态内容特征图对风格特征模型中的风格特征进行学习;最后使用迁移学习TL（transfer learning）获得步态虚拟偏移样本。实验结果表明,经过DLTL风格学习生成的步态虚拟样本虽然整体风格发生改变,但人体步态特征没有改变,可有效扩充小样本容量;当虚拟样本增加到一定数量时,步态识别率有所提升。该方法与现有步态虚拟样本生成方法进行对比实验,结果表明该算法优于现有方法,能够大量生成虚拟样本且稳定提高步态识别的识别率。     

基于核二维主成分分析算法的步态识别

步态识别是一种新的生物认证技术,它是通过人的行走方式来识别人类身份的方法。为了更加快速有效地对人体步态特征进行提取和识别,采用了基于核二维主成分分析（Kernel two Dimensional Principal Component Analyses,K2DPCA）的方法进行步态特征提取,运用支持向量机（SVM）进行步态识别。根据人体步态下肢摆动距离统计出步态周期,得到步态能量图（GEI）,对生成的GEI采用核二维主成分分析方法进行步态特征向量提取,采用SVM分类器进行分类识别。实验结果表明该方法具有很好的识别效果。     

基于下肢角度特征的步态识别方法

基于人行走时的下肢角度变化包含丰富的个体识别信．幽观点,提出利用下肢角度特征进行步态识别的新方法。对每个步态序列,依据人体解剖学的先验知识定位下肢关节点,计算相邻关节点连线与竖直线的夹角,以此作为下肢角度;通过步态周期分析,提取一个步态周期的下肢角度变化序列作为特征向量表征步态。最后,采用针对小样本问题具有很好分类效果的支持向量机技术实现步态的分类决策。CASIA步态数据库上的仿真结果证明本方法具有较高的识别性能。     

神经网络滑模控制下肢外骨骼机器人的轨迹跟踪

针对下肢外骨骼机器人行走稳定性与步态轨迹跟踪控制问题,对下肢外骨骼机器人三连杆模型进行动力学建模与轨迹仿真。通过拉格朗日法建立下肢外骨骼机器人的动力学模型,设计了神经网络自适应滑模控制算法。引入神经网络,对下肢外骨骼机器人步态轨迹跟踪系统的不确定项进行逼近,在控制器中采用了改进的趋近律,使用李雅普诺夫稳定性理论进行了稳定性分析,并通过MATLAB对改进后的控制算法进行了仿真验证。仿真结果表明,采用该算法对具有关节摩擦和外界环境干扰的下肢外骨骼机器人进行轨迹跟踪时,具有较好的跟踪效果;通过改进的趋近律,能削弱系统的抖振。相比于基于计算力矩法的滑模控制,该控制算法有更好的跟踪效果,能应用到下肢外骨骼机器人行走的稳定性和步态轨迹跟踪控制中。     

基于有监督Kohonen神经网络的步态识别

表面肌电信号随着时间的变化而改变,这将影响运动模式的分类精度.传统人体下肢假肢运动模式的识别算法不能保证在整个肌电控制时间内达到对运动模式的有效识别.为了解决这些问题,本文提取步态初期200ms的信号的特征值,将无监督和有监督的Kohonen神经网络算法应用到大腿截肢者残肢侧的步态识别中,并与传统BP神经网络进行了对比.结果表明,有监督的Kohonen神经网络算法将五种路况下步态的平均识别率提高到88.4%,优于无监督的Kohonen神经网络算法和BP神经网络.     

结合注意力卷积网络与分块特征的步态识别

目前深度学习算法已经广泛应用于步态识别领域,但是大多数现有方法通过卷积神经网络提取步态全局特征时,忽略了许多包含关键步态信息的局部特征,在一定程度上削弱了步态识别的精度和提升潜力。针对上述问题,提出了一种结合注意力卷积神经网络与分块特征的跨视角步态识别方法,该方法以步态轮廓图序列为输入,每帧图片分别经过相同结构的注意力卷积神经网络融合成整体特征,在网络中加入有效的注意力机制CBAM能显式地建模各空间及通道的重要程度,增大显著区域特征的权重;整体特征被水平分成两块进行训练和步态识别,提取的步态局部特征更适合精细的步态分类。在步态数据集CASIA-B和OU-ISIR-MVLP上进行跨视角步态识别实验,结果证明在训练数据集充足与不足的条件下,该方法在识别精度上均优于现有方法。     

Neural computing for walking gait pattern identification based on multi-sensor data fusion of lower limb muscles

The use of neural computing for gait analysis widely known as computational intelligent gait analysis is addressed recently. This research work reports multilayer feed-forward neural networks for walking gait pattern identification using multi-sensor data fusion; electromyography (EMG) signals and soft tissue deformation analysis using successive frames of video sequence extracted from lower limb muscles according to each gait phase within the considered gait cycle. Neural computing framework for walking gait pattern identification consists of system hardware and intelligent system software. System hardware comprises a wireless surface EMG sensor unit and two video cameras for measuring the neuromuscular activity of lower limb muscles, and a custom-developed artificial neural network for classifying the gait patterns of subjects during walking. The system uses root mean square and soft tissue deformation parameter as the input features. Multilayer feed-forward back propagation neural networks (FFBPNNs) with different network training functions were designed, and their classification results were compared. The intelligent gait analysis system validation has been carried out for a group of healthy and injured subjects. The results demonstrated that the overall accuracy of 98 % prediction is achieved for gait patterns classification established by multi-sensor data fusion of lower limb muscles using FFBPNN with Levenberg–Marquardt training function resulting better performance over FFBPNN with other training functions.

     

Multi-channel Fusion Based Adaptive Gait Trajectory Generation Using Wearable Sensors

This paper presents a method to regenerate lower limb joint angle trajectories during gait cycle by judging human intention using wearable sensor system. Myoelectric signals from user are used to detect the intention of gait initiation and gait phases. Multi-channel redundant fusion technique is implemented to obtain a robust stride time and gait phase calculation algorithm. Joint trajectories corresponding to particular gait events and phases are regenerated using a Radial basis neural network. The network is trained with joint angle data measured by Inertial Measurement Unit (IMU) from users with varying anthropomorphic features. Generated trajectory is adaptive to anthropomorphic as well as gait velocity variation. Contribution of this paper is in development of a wearable sensor system, multi-channel redundant fusion to calculate stride time and an adaptive gait trajectory generation algorithm. The proposed method of trajectory generation is used to regenerate lower limb joint motion in sagittal plane for wearable robotic devices like prosthesis and active lower limb exoskeleton.     

基于多支路残差深度网络的跨视角步态识别方法

针对基于卷积神经网络的步态识别模型不能充分利用局部细粒度信息的问题,提出基于多支路残差深度网络的跨视角步态识别方法.将多支路网络引入卷积神经网络中,分别提取步态轮廓序列图中不同粒度的特征,并利用残差学习和多尺度特征融合技术,增强网络的特征学习能力.在公开步态数据集CASIA-B和OU-MVLP上的实验证实文中方法的识别准确率较高.     

基于视角转换的多视角步态识别方法

针对步态识别中步态视角变化、步态数据样本量少及较少利用步态时间信息等问题,提出一种基于视角转换的步态识别方法。通过VTM-GAN网络,将不同视角下的步态能量图及含有步态时间信息的彩色步态能量图,统一映射到保留步态信息最丰富的侧视图视角,以此突破步态识别中多视角的限制,在视角转换的基础上,通过构建侧视图下的步态正负样本对来扩充用于网络训练的数据,并采用基于距离度量的时空双流卷积神经网络作为步态识别网络。在CASIA-B数据集上的实验结果表明,该方法在各状态、各角度下的平均识别准确率达到92.5%,优于3DCNN、SST-MSCI等步态识别方法。     

Human lower extremity joint moment prediction: A wavelet neural network approach

Joint moment is one of the most important factors in human gait analysis. It can be calculated using multi body dynamics but might not be straight forward. This study had two main purposes; firstly, to develop a generic multi-dimensional wavelet neural network (WNN) as a real-time surrogate model to calculate lower extremity joint moments and compare with those determined by multi body dynamics approach, secondly, to compare the calculation accuracy of WNN with feed forward artificial neural network (FFANN) as a traditional intelligent predictive structure in biomechanics.To aim these purposes, data of four patients walked with three different conditions were obtained from the literature. A total of 10 inputs including eight electromyography (EMG) signals and two ground reaction force (GRF) components were determined as the most informative inputs for the WNN based on the mutual information technique. Prediction ability of the network was tested at two different levels of inter-subject generalization. The WNN predictions were validated against outputs from multi body dynamics method in terms of normalized root mean square error (NRMSE (%)) and cross correlation coefficient (ρ).Results showed that WNN can predict joint moments to a high level of accuracy (NRMSE < 10%, ρ > 0.94) compared to FFANN (NRMSE < 16%, ρ > 0.89). A generic WNN could also calculate joint moments much faster and easier than multi body dynamics approach based on GRFs and EMG signals which released the necessity of motion capture. It is therefore indicated that the WNN can be a surrogate model for real-time gait biomechanics evaluation.     

Realization of Dynamic Walking and Running of the Quadruped Using Neural Oscillator

In the present study we attempt to induce a quadruped robot to walk dynamically on irregular terrain and run on flat terrain by using a nervous system model. For dynamic walking on irregular terrain, we employ a control system involving a neural oscillator network, a stretch reflex and a flexor reflex. Stable dynamic walking when obstructions to swinging legs are present is made possible by the flexor reflex and the crossed extension reflex. A modification of the single driving input to the neural oscillator network makes it possible for the robot to walk up a step. For running on flat terrain, we combine a spring mechanism and the neural oscillator network. It became clear in this study that the matching of two oscillations by the spring-mass system and the neural oscillator network is important in order to keep jumping in a pronk gait. The present study also shows that entrainment between neural oscillators causes the running gait to change from pronk to bound. This finding renders running fairly easy to attain in a bound gait. It must be noticed that the flexible and robust dynamic walking on irregular terrain and the transition of the running gait are realized by the modification of a few parameters in the neural oscillator network.