A Survey on Deep Learning-Based 2D Human Pose Estimation Models

In this article,a comprehensive survey of deep learning-based(DLbased)human pose estimation(HPE)that can help researchers in the domain of computer vision is presented.HPE is among the fastest-growing research domains of computer vision and is used in solving several problems for human endeavours.After the detailed introduction,three different human body modes followed by the main stages of HPE and two pipelines of twodimensional(2D)HPE are presented.The details of the four components of HPE are also presented.The keypoints output format of two popular 2D HPE datasets and the most cited DL-based HPE articles from the year of breakthrough are both shown in tabular form.This study intends to highlight the limitations of published reviews and surveys respecting presenting a systematic review of the current DL-based solution to the 2D HPE model.Furthermore,a detailed and meaningful survey that will guide new and existing researchers on DL-based 2D HPE models is achieved.Finally,some future research directions in the field of HPE,such as limited data on disabled persons and multi-training DL-based models,are revealed to encourage researchers and promote the growth of HPE research.

Gray Model of Human Body Shape

In order to determine the control dimensions of human body in pattern construction and provide the basis for size specification in clothing industry, a gray model is described and established in the application of predicting size specification. By comparing the features of gray model and regression model, it is proved that the former has many advantages over the latter, i.e. more practical and more effective in this application. So it is strongly suggested to apply the innovative gray model instead of the traditional regression model to facilitate the calculation procedure in pattern construction.

Rapid 3D human body modeling based on Kinect technology

The 3D shape of the human body is useful for applications in fitness, games and apparel. Accurate body scanners, howev- er, are expensive, limiting the availability of 3D body models. In this paper, we focus on the human body 3D modeling using one Kinect sensor modeling method. To get the human body model fast, three steps in rapid modeling of the human body are carried out. Firstly, according to anthropometric parameters, the standard model is parameterized; Secondly, the Kinect depth image of the human body model is gotten through, then, by using the PCL library the point cloud data is processed and matched, and the human body model is optimized; Finally, the realistic human body model is obtained with the rapid integration of the standard model and PCL library.

Research on Dynamic Model of the Human Body

After summarizing the current situation of the research on human body modeling, a new dynamic model containing 5 equivalent masses has been proposed and the corresponding dynamic equations has been deduced too. By using this new model, more detailed information about the situation of the human body under impact and vibration can be obtained. The new model solves the problem that transmission functions of forces inside the human body can't be deduced by using 3-equivalent-mass model. It will find its usage in many applications.

Kinematic simulation of human gait with a multi-rigid-body foot model

The paper builds a multi-rigid-body model of human with a 4-rigid-body foot in the 3D CAD software Solidworks, based on human anatomy. By controlling the rotation of the ankle and major joints of human body while walking, the Kinematic simulation was performed in the dynamics simulation software ADAMS. The paper analyzes the simulate results and points out deficiencies in the current work and the direction of research efforts in future.

Human Body Modeling and Posture Simulating Based on 3D Surface Scan Data

This paper presents a new approach for modeling the human body by considering the motion state and the shape of whole body. The body model consists of a skeleton kinematic model and a surface model. The former is used to determine the posture of the body,and the latter is used to generate the body shape according to the given posture. The body surface is reconstructed with multi-segment B-spline surfaces based on the 3D scan data from a real human body.Using only a few joints parameters and the original surface scan data, the various body postures and the shape can be generated easily. The model has a strong potential of being used for ergonomic design,garment design, virtual reality environment, as well as creating human animation, etc.

Research on Modeling of Human Body in Three-Dimensional Computer Animation

Modeling of the human body is a basic problem in computer animation of the humanbody. This paper presents a method in which the combination of polyhedra is used to establish aninitial rough design of the human body and subdivision of polyhedra to approximate the curvesurface of the real human body. In addition, the dynamic characteristics are taken intoconsideration of the human body, thus providing a fine basis for the establishment of the bumanbody animation.

The two-way feedback and passing-way of human body

Two-way feedback of human body was published in 1992. The sensation of two-way feedback of body is a special system of human reaction, which maintains and regulates symmetry and balance of human body. The human two-way feedback reacts to human health. For human overall health and delay decrepitude, it is necessary to pay attention to the stimulations (passive acceptance and initiative interventions) and relevant influences in human body and the stimulative effect. In this paper, the experimental research of stimulation and an example of two-way feedback in human body are given. And lay a foundation of prevention, medical treatment and hygiene of human overall health.

HBM、CDM静电放电模型及其失效特征研究

本文介绍了人体模型、带电器件模型两种典型的静电放电模型,对人体模型、带电器件模型的区别、静电放电失效机理进行了说明。通过对芯片进行人体模型、带电器件模型静电放电极限测试,获取不同静电放电模型芯片失效电压及其失效特征。实验结果表明,芯片抵抗人体模型、带电器件模型静电放电的能力没有直接关系,并不是抵抗人体模型静电放电能力强,耐带电器件模型静电放电能力就高;人体模型、带电器件模型静电放电极限测试后芯片损伤位置、损伤形貌存在差异。

基于2D人体图像特征学习的女西装合体性判别

为解决服装远程定制和网络购衣过程中服装合体性无法预先判断的问题,文章提出利用2D人体图像判断女西装合体性。首先采集了462名18~25周岁青年女性的2D图像及3D尺寸,建立人体数据集。并制作10件虚拟女西装,建立服装数据集。其次提取用于女西装合体性判别的服装特征、2D人体图像关键点距离特征和整体特征。基于感性工学原理,采用专家评价法获取女西装合体性评价标签。最后利用贝叶斯分类器建立女西装合体性判别模型。研究结果显示,将文章提出的特征应用于基于贝叶斯算法的女西装合体性判别模型判别准确率可达到84.8%,说明提出的服装合体性评价指标有效,可以用于基于2D图像的服装合体性预测。

移乘背抱护理机器人舒适性生物力学模型分析

为探索移乘背抱护理机器人人机交互力学特性与被护理人舒适性感受的潜在关联,研究显性的人机接触力相对于人体生物力学特性的映射关系,揭示影响被护理人舒适性的主要作用肌群,提出了舒适性测试试验与AnyBody背抱运动人体生物力学模型相结合的方法.从被护理人的肌肉激活程度入手,对移乘背抱护理机器人背抱运动舒适性特征进行了分析.首先,介绍了移乘背抱护理机器人典型结构工作原理并进行了运动学分析,在此基础上搭建了舒适性测试试验系统,基于主观舒适性评分规划了服务于试验的基础运动轨迹,并通过试验得到了人机主要接触部位力学信息.然后,以试验数据为边界条件构建了基于AnyBody的背抱运动人体生物力学模型,依托逆动力学仿真分析得到了肌肉激活程度,并对比舒适性测试肌肉激活程度验证了仿真模型的有效性.最后,结合人机主要接触部位受力、肌肉激活程度及主观舒适性评分综合分析了移乘背抱护理机器人背抱运动舒适性特征,明确了影响被护理人舒适性的主要作用肌群.结果表明:影响被护理人舒适性的主要作用肌群有7个,且被护理人胸部压力与腋下压力值大小分别与影响其躯干和上肢舒适性的肌肉激活程度呈正相关.可通过建立胸部压力与腋下压力值相关的加权函数有效描述不同被护理人整体舒适性特征,为移乘背抱护理机器人舒适性运动控制提供了理论依据.

基于虚拟仿真技术的服装智能制造应用研究

由于传统的服装制造不能很好地满足市场需求,因此对虚拟仿真技术在服装智能制造中的应用进行研究,实现服装企业转型升级,确保企业健康发展。服装虚拟设计包括仿真人体建模、虚拟服装建模、3D仿真优化和构建应用平台。采用非接触式测量获取人体三维数据,构建虚拟三维人体模型,同时将服装设计师的设计意图进行深层次挖掘,获得三维虚拟服装模型。通过纹理特征映射、碰撞检测、真实感增强和场景构建来优化服装设计过程,并将其应用于服装个性化设计与虚拟试穿中,对提升服装企业的智能制造水平和市场竞争力具有一定的参考价值。

基于单-多视图优化的足球球员三维姿态和体型估计

足球比赛场景的三维重建有助于观众自由切换视角,增加了互动性和沉浸感。针对足球比赛场景中的足球球员,提出一种三维姿态和体型估计方法。对球员的多视图图像使用训练好的部分注意力回归的三维人体估计(PARE)模型生成初始的三维姿态和体型估计,并使用人工标注的二维关节点作为优化目标。单-多视图优化操作利用蒙皮多人线性模型(SMPL)和正交投影的可微性,将球员的三维姿态和体型参数映射到二维关节点,计算其与人工标注之间的差异,再使用神经网络的反向传播算法更新三维姿态和体型参数,持续这些过程直到差异最小化。在自建的足球球员多视图数据集上的实验结果表明,该方法能够有效估计足球球员的三维姿态和体型,与人体网格恢复、在循环中优化SMPL、PARE等方法相比,二维关节点精度在单视图上提高了9.2%~37.5%,在多视图交叉验证中提高了34.9%~54.1%。

基于人体振动试验的坐姿人体模型研究

为了能够准确模拟人体受振特性,基于多体动力学原理,将人体简化成头部、上躯干、下躯干和腿臀部4集总质量、12自由度多刚体模型,模型中考虑上躯干、下躯干和腿臀部之间的腰部肌肉作用,各部分之间采用线性等效弹簧和阻尼进行连接。借助振动试验台,对8名志愿者进行了坐姿低频振动试验,得到了人体在垂向激励(0~20 Hz)和纵向激励(0~15 Hz)下座椅-头部的振动传递特性曲线。模型中模拟人体动态特性的48个刚度、阻尼参数,采用自适应模拟退火算法对试验得到的座椅-头部振动传递特性曲线进行参数辨识来确定。结果显示,12自由度多刚体动力学人体模型仿真曲线与试验曲线相一致,模型能够较为准确反映人体受振特性。

智能可穿戴监测服装性能评价模型的研究进展

为促进可穿戴智能服装评价标准形成进而促进可穿戴智能服装产业化发展,系统介绍人体-织物数理模型及人体与织物间动态压力数值模拟的评价模型。基于弹性力学理论,探究静止状态下拉普拉斯数理模型的研究进展;介绍织物结构对接触压力的影响及织物结构的数值模拟,并进一步探究着装过程、运动状态下人体-织物间连续动态压力的数值模拟情况,以及数值模拟法与计算机技术结合的现状。最后讨论现阶段可穿戴智能服装评价模型的应用,并提出使用数值模拟法进行多物理场耦合、模拟加速度运动用于提高可穿戴智能服装可靠性评价标准的发展展望。

下肢肌力的跌倒风险预测的方法研究

分析人体行走和跌倒时下肢肌肉的作用,确定对行走稳定性有主要影响的肌肉,为跌倒预测和信号采集位置提供依据。首先建立了一个人体骨骼肌模型,并使用Vicon三维红外动作捕捉分析系统和AMTI三维测力台采集记录模拟行动不便人群的步态、向后摔倒和向前摔倒的实验数据;然后运用逆运动学分析、静态优化算法,通过OpenSim仿真软件计算出步行和跌倒时下肢冗余肌力分配;最后对数据进行分析。研究结果表明:通过仿真模型定量分析了步行和跌倒过程中下肢肌力分布,量化步行以及跌倒过程中下肢各单块肌肉力量的变化规律,得出采集肌电信号较为合适的位置为股外侧肌和比目鱼肌,并为预测跌倒以及康复指导提供了生物力学依据,为进一步揭示髋、膝、踝关节的生物力学特性和人体下肢康复机器人的设计奠定基础。

面向人体动作视频的语义编码传输方法

提出了一种高效的人体视频的无线传输方法,称为人体动作视频语义传输(AAST)系统。在当今的应用场景中,H.265视频编码和5GLDPC信道编码的组合已成为视频传输的常见方法。然而,在传输效率方面,尤其是在特殊情况下,仍存在相当大的提升潜力。受到最近人体动作视频研究的进展启发,提出了AAST系统。AAST系统能够以高效的方式传输关键区域的语义信息,并在接收端重建原始视频。鉴于关键区域运动语义信息内在的相关性,AAST系统采用非线性分析变换将运动信息映射到潜在空间。然后通过信源信道联合编码,将运动信息的潜在表示传输到接收端。在该系统中,利用潜在表示的先验信息来估计关键区域语义信息的重要性。这有助于在信源信道联合编码中实现自适应码率控制,从而带来显著的编码增益。大量的实验证明了AAST系统相对于其他有竞争性的视频传输系统具有更卓越的性能。值得注意的是,与H.265视频编码和5GLDPC信道编码传输方法相比,在达到相同的感知指标时,AAST系统最高可以节省约70%的信道带宽,展现了其在视频无线传输中的鲁棒性。

基于多监督的三维人体姿势与形状预测

三维人体重建技术指通过图像或视频建立有相应姿势和体型的三维人体模型,其在虚拟现实(VR)、网游、虚拟试衣等方面有着十分广泛的应用前景.其中,参数化的三维人体模型由于参数数量的局限,重建精度较低,缺少细节特征.为了提升参数化三维人体模型的重建精度,增加其脸部与手部细节,提出一种基于多监督的三维人体模型重建方法.该方法结合传统的回归方法和优化方法,利用卷积神经网络回归出参数化人体模型的参数,得到一个较为粗糙的人体模型,将该模型作为初始模板进行拟合和迭代优化,将带有脸部和手部的全身密集关节点信息和轮廓信息作为回归网络的2D监督,同时使用迭代优化后的人体模型作为回归网络的3D监督,最终可由一幅图像获得一个多细节、高精度的参数化三维人体模型.定性分析结果表明,该方法为人体拟合过程提供正确的拟合方向,可有效减少非自然姿势的出现,提高三维人体模型重建的准确度.全身密集关节点监督可为模型增加更多手部与脸部的细节,而轮廓监督可减少重建的人体模型与图像中人体的像素级偏差.定量分析表明,该方法在数据集Human3.6M上的平均逐关节位置误差(MPJPE)为59.9 mm,较经典方法SPIN减少了4.16%,对关节点进行刚性对齐后模型的平均逐关节位置误差(MPJPE-PA)低至38.2 mm,较SPIN减少了7.06%.

IC器件在HBM ESD脉冲电压下的介质击穿实验研究

在IC器件氧化绝缘层介质击穿物理模型的基础上,讨论了人体带电放电模型(HBM ESD)中的波动电压 加于氧化层上时的介质击穿．发现了氧化层上的俘获空穴密度临界值随薄膜厚度的变化十分敏感,当厚度由 0．010 μm减少至0．003 μm时,俘获空穴密度临界值由0．100 C／cm2减少至0．002 C／cm2．相应地,IC器件抗ESD 的性能也急剧下降．