Features of the Postural Sway Signal as Indicators to Estimate and Predict Visually Induced Motion Sickness in Virtual Reality

Navigation in a 3D immersive virtual environment is known to be prone to visually induced motion sickness (VIMS). Several psychophysiological and behavioral methods have been used to measure the level of sickness of a user, among which is postural instability. This study investigates all the features that can be extracted from the body postural sway: area of the projection of the center of gravity (mainly considered in past studies) and its shape and the frequency components of the signal’s spectrum, in order to estimate and predict the occurrence of sickness in a typical virtual reality (VR) application.

After modeling and simulation of the body postural sway, an experiment on 17 subjects identified a relation between the level of sickness and the variation both in the time and frequency domains of the body sway signal. The results support and go further into detail of findings of past studies using postural instability as an efficient indicator of sickness, giving insight to better monitor VIMS in a VR application.     

基于步态的身份识别

提出了一种简单有效的自动步态识别算法，对于每个序列而言，一种改进的背景减除方法用于检测行人的运动轮廓，然后，这些时变的2D轮廓形状被转换为对应的1D距离信号，同时通过特征空间变换来提取低维步态特征，基于时空相关或归一化欧氏距离度量，标准的模式分类技术用于最终的识别，实验结果表明该算法不仅获得了令人鼓舞的识别性能，而且拥有相对较低的计算代价。     

基于步态的身份识别综述

提出了步态识别的重要意义,介绍了基于步态的身份识别的过程。对步态识别的国内外研究现状和研究内容做了概述,列举了常用的几个步态数据库,并阐述了用在步态识别各部分的一些经典方法。对该课题的后续工作作了展望。     

Dynamic Studies On Human Body Sway By Using A Simple Model With Special Concerns On The Pelvic And Muscle Roles

The human pelvis is such a unique structure that enables our upper body to work so perfectly with the two legs so as to control the body's balance in the complicated postures. The aim of this study is to establish a new dynamic body sway control model in the upright standing body position in coronal plane, and to reveal the possible control mechanisms underlying the body sway with special concerns on the roles that the pelvis and its muscles are performing during the sway. The plant of control model, the dynamics of human body, includes five parts, i.e. two ankles, two hips and one lumbosacral joint, which makes up a multi‐link inverted pendulum system, and is driven by two pairs of muscles, the psoas major (PM) and glutaeus medius (GM). Body sway records from eight healthy young subjects showed that the angular sway scopes of the ankle on roll (lateral) plane are 0.94±0.36± (eye‐open) and 1.35±0.52± (eye‐closed) respectively, while in lumbosacral plane, the scopes are 0.99±0.41± (eye‐open) and 1.27±0.72± (eye‐closed). The ankle and lumbosacral sways were almost in the same degree, yet their phase difference was near ±n, which means that the body trunk maintains perpendicular to horizon during the upright stance. Surface electromyographic (sEMG) activity from GM also showed the same evidence: the activated GM was always in the same side as the deviated center‐of‐pressure (COP). By assuming the corrective torque of posture is regulated by PID (proportional, integral and derivative) control, the body sway can be simulated by applying human physical parameters. Our study results demonstrated that the simulated traces are consistent with the experimental recorded, suggesting that the pelvis is an important structure for the posture maintenance and control, and the mechanism of balance keeping control during upright stance can be approximately taken as a PID control. The result also suggests a novel means for postural stability assessment in individual in the future.     

基于步态序列图像的身份确认

为了利用HMM抽取的步态序列的动态特征来进行身份确认，首先提出一种改进的角度向量用来表征二值化的步态序列图像，以便将每幅图像转化为1维向量，然后再以此作为特征向量，对每个人物建立并训练HMM模型，用于确定人物身份。这种改进的角度向量由于具有较强的抗噪性和方便的尺度伸缩性能，因此既适用于分割质量较差的图像，又能减小行走方向和距离的影响。实验表明，这种HMM不仅能较好地模拟步态的动态特征，还能描述序列图像间的联系，而且算法执行速度快，从输入原始数据到输出识别结果所需时间不超过2min，能满足实时要求。在Soton和NLPR数据库上进行的实验，分别获得了100％和85％的识别率，证明该方法是有效的。     

基于心电图的身份识别方法

针对当前心电图(ECG)身份识别中存在的小样本、多特征点检测问题,提出基于小波变换和动态时间规整(DTW)相结合的方法.利用小波变换对ECG信号进行预处理并提取R波峰值点,提取并保存肢导联QRS波及心拍模板,根据QRS波测试数据与各QRS波模板间的相关性分析以及阈值条件缩小身份识别范围,采用 DTW算法确定心拍测试数据...     

Postural analysis of nursing work

Back pain in the nursing profession is an acknowledged wide spread occupational hazard. This study used OWAS (Ovako Working posture Analysis System) to measure the severity of the working postures adopted by nurses on Care of the Elderly wards when carrying out manual handling operations for animate and inanimate loads. Twenty-six nurses were observed on 31 occasions to obtain 4299 observations, these data were collected and processed using the OWASCO and OWASAN programs, and then analysed by grouping the results into defined patient (animate) handling and non-patient (inanimate) handling tasks. A statistical comparison was made between the two groups using the percentage of action categories two, three and four, to the total number of action categories. A significant difference (p < 0.05) was found, demonstrating that the percentage of harmful postures adopted during patient handling tasks was significantly higher than during non-patient handling tasks. This high level of postural stress and the poor track record of risk management within the Health Care Industry leads to the recommendation that an attitudinal change is needed to successfully address and reduce the manual handling burden which is currently being carried by the nursing staff.     

Postural Control as a Probe for Cognitive State: Exploiting Human Information Processing to Enhance Performance

The battlefield of the future will require the warfighter to multitask in numerous ways, seriously taxing the cognitive and perceptual capabilities of even the most advanced warrior. A principal concern in developing a better understanding of how current and proposed computational technologies can supplement and augment human performance in this and other environments is determining when such assistance is required. This challenge can be parsed into 2 components: determining what set of measurements accurately reflects cognitive state, and identifying techniques for synthesizing this set of measurements into a single collective cognitive state variable. The primary thesis of this proposal is that automatic human behavioral responses serve as inherent probes for cognitive state. Further, the human perception-action system is uniquely designed to capture, process, integrate, and act on an extraordinarily diverse range of information freely available in the natural environment. Together, this system and the surrounding environment which acts on it-and on which the system acts-form a dynamic coupling. Under normal conditions these couplings remain intact. When stressed, these couplings become degraded. Based on this understanding, the authors propose a unique suite of Cognitive Workload Assessment (CWA) tools, based on real-time measurements of postural control that can serve as both a stand-alone indicator of cognitive state as well as a cueing filter for engaging other CWA sensor suites that are currently under development.     

基于人体身型曲线特征的身份鉴定技术

研究人体身型特征的身份识别,提高身份准确性问题。针对传统的人体身份识别技术,都是用直线距离近似表示人体身型的几何特征,但是人体身型中的大部分真实几何特征距离都是空间曲线距离,会导致测量的身型几何特征结果不够精确,造成后期的身份识别准确性不高的问题。为了解决上述问题,提出一种基于近似曲线几何特征的身份鉴定方法,运用近似曲面分割逼近技术,求出人体身型几何特征点之间的最短近似路径,运用多个最短路径逼近方法,准确获得人体真实的身型几何特征信息。实验结果表明,改进的方法能够准确提取三维人体身型几何特征,使得身份识别准确率大幅增加。     

人体运动的检测和识别研究

本文针对人体运动视觉分析中的行为理解和分析等高层视觉问题进行分析,研究了一种静止摄像机条件下的行为理解和分析的算法,它以运动序列中的关键帧为基础,针对关键帧提取人体的骨架信息,然后通过Hu不变矩来提取特征,最后组成特征向量,通过对HMM模型的训练来识别特定运动序列的语义.     

一种人手背静脉特征识别方法

针对目前手背静脉图像识别采用细化和骨架操作等提取结构特征易造成静脉结构细节丢失和特征点误判等问题,提出一种基于方向梯度直方图(HOG)的手背静脉特征识别方法。采用生物特征识别的一般流程,对手背静脉图像灰度进行归一化和滤波增强等预处理后,直接对手背静脉灰度图像进行二级小波包分解,提取低频子带图的HOG纹理特征,最后采用K近邻分类器实现个人身份识别。利用自行建立的手背静脉图像数据库对所提方法进行验证,结果证明了算法的有效性,其正确识别率为95%,应用前景广阔。     

基于步态特征的快速身份识别方法

步态是远距离情况下人的身份识别的一个重要的特征,论文提出了一种基于步态特征的快速身份识别方法。用人体的宽度特征来分析步态运动,提取关键帧,对关键帧进行人体轮廓的提取,运用傅立叶描述子提取模板。最后用最近邻法进行匹配,实现人的身份识别。     

多人体识别监控技术的研究

多人体识别监控指的是以图像处理、模式识别、计算机视觉等技术为基础,对某一区域的人群进行识别监控.文章从数字图像处理技术的发展和目前国内外研究现状进行分析,对多人体识别技术中运用的图像分割、处理和特征值提取的主要方法进行了总结,对多人体识别监控技术存在的问题及发展方向进行了讨论.     

Postural stress analysis in industry

Both observational and instrumentation-based techniques have been used to conduct postural stress analysis in industry. As observational methods are more widespread than instrumentation-based techniques and can be used as a practical tool in the workplace, this study reviews and assesses the scientific literature on observational techniques. Techniques are classified into macropostural, micropostural and postural-work activity. The basis for each classification is outlined and evaluated. Postural recording is performed either continuously or intermittently. Intermittent postural recording procedures lack the criteria for determining the optimum number of observations for low and high repetitive jobs. Research is warranted to examine the sources and magnitudes of errors associated with postural classification. Such information is required to train job analysts in the ergonomics of working postures.     

减摇水舱试验台架横荡系统研究

船舶减摇水舱试验台架是研究和设计减摇水舱的重要实验设备,准确模拟船舶在海浪中的运动是设计试验台架的关键。目前试验台架研究的仅是单纯横摇,如果考虑了横荡的基本轨圆运动,以及相应的横摇与水舱内液体加速度的耦合,则船的横摇运动更符合实际情况．本文对横荡运动进行了理论分析,并将PID控制方法应用于模拟横荡运动的电液位置伺服控制系统中,仿真结果表明系统具有良好的跟踪性能．符合设计要求．     

基于Wi-Fi信号的人体身份识别算法研究综述

近年来,Wi-Fi感知凭借低成本、非接触式、不受光照影响、隐私性更好等优势,成为人机交互的新兴研究方向.目前,Wi-Fi感知研究已从目标定位扩展到动作识别、身份识别等领域.以人体身份识别为例,对Wi-Fi感知技术在该领域的研究进行了总结和分析.首先,对Wi-Fi感知技术的发展历史及优缺点进行了简要概述,并介绍了与传统的身份识别技术相比,利用Wi-Fi信号进行身份识别的优势;其次,详细介绍了Wi-Fi感知身份的基本流程,其中包括信号采集、预处理、特征提取、身份识别4个步骤,并具体介绍了每个步骤具体的操作过程;然后,从步态与手势两个方面对现有的Wi-Fi感知身份研究成果进行了横向与纵向的对比和分析;最后,针对目前该领域研究中存在的关键问题,提出了未来的研究重点,主要包括多人身份识别与迁移学习.     

基于步态特征的身份识别算法研究

步态识别根据人走路的姿势进行身份识别,由于人在行走时在空间呈现出的不同几何模式,单一几何特征难以全面描述步态特征,导致身份识别正确率不高。为提高身份识别的正确率,提出一种空间和频率特征模式相融合的身份识别算法。首先利用摄像机采集步态图像序列,然后分别采用极坐系和傅里叶变换提取步态空间特征和频率特征,并对两种特征进行融合,最后采用支持向量机对融合特征进行学习和分类,进行身份识别。结果表明,相对于单一步态特征为参数的身份识别算法,融合算法的身份识别正确率有了明显提高,且具有更好的稳定性。     

脉冲输入下基于输入整形法的起重机防摇摆方法

起重机在工业生产中有着广泛的应用,在其装卸货物的过程中,吊重产生的摇摆,会影响码头的安全性及运输效率。所以提出了一种适合工业环境,成本最优,摆角回零时间最优的起重机防摇摆的方法。首先,建立了吊重的动力学模型,经过分析加速度输入和摆角输出后,提出了脉冲加速度输入下一族基于输入整形法的起重机防摇方法。在不同时刻,让不同幅值的脉冲响应进行叠加,当输入为二脉冲,三脉冲,四脉冲,......,n脉冲时,摆角输出能在1/2,2/3,3/4,......,(n-1)/n个系统震荡周期内回到零。最后,以MATLAB为仿真平台,对该方法进行仿真,仿真结果表明,此控制方法具有良好防摇的效果。