Modeling the Influences of Cyclic Top-Down and Bottom-Up Processes for Reinforcement Learning in Eye Movements

Understanding and reproducing complex human oculomotor behaviors using computational models is a challenging task. In this paper, two studies are presented, which focus on the development and evaluation of a computational model to show the influences of cyclic top-down and bottom-up processes on eye movements. To explain these processes, reinforcement learning was used to control eye movements. The first study showed that, in a picture-viewing task, different policies obtained from different picture-viewing conditions produced different types of eye movement patterns. In another visual search task, the second study illustrated that feedback information from each saccadic eye movement could be used to update the model's eye movement policy, generating different patterns in the following saccade. These two studies demonstrate the value of an integrated reinforcement learning model in explaining both top-down and bottom-up processes of eye movements within one computational model.     

Movements with Lateral Constraint

Abstract

The literature on terminal control tasks and continuous tracking tasks suggests a possible model for tasks where the subject must guide a vehicle at his own speed along a course defined by a tolerance band. The model predicts a linear relationship between the average velocity and the width of the tolerance band. Throe experiments are described which simulate vehicle guidance by pencil-and-paper experiments for courses consisting of straight lines and circles. The predictions of the model were confirmed, suggesting that a general measure for the ease of control of a vehicle (or other controlled system) is possible using this technique.     

Example-Based Rendering of Eye Movements

This paper describes a model for example-based, photo-realistic rendering of eye movements in 3D facial animation. Based on 3D scans of a face with different gaze directions, the model captures the motion of the eyeball along with the deformation of the eyelids and the surrounding skin. These deformations are represented in a 3D morphable model.
Unlike the standard procedure in facial animation, the eyeball is not modeled as a rotating 3D sphere located behind the skin surface. Instead, the visible region of the eyeball is part of a continuous face mesh, and displacements of the iris as well as occlusions by the lids are modeled in a texture mapping approach. The algorithm avoids artifacts that are widely encountered in 3D facial animation, and it presents a new concept of handling occlusions and discontinuities in morphing algorithms.     

Estimation of Human Movements via Quaternion

This paper presents the authors＇ research results about HRI （human-robot interaction）. The goal is to estimate together the arm position, the anatomical movements of the shoulder and accelerations of the arm with respect to the shoulder and visualizing this movement in a 3-D virtual model to control a Robot. The estimation algorithm makes use of a nonlinear observer and an optimization routine to fuse information from the sensor. The global asymptotic convergence of the nonlinear observer is guaranteed. Extensive tests of the presented methodology with real world data show the effectiveness of the proposed procedure.     

Learning Physically Based Humanoid Climbing Movements

We propose a novel learning‐based solution for motion planning of physically‐based humanoid climbing that allows for fast and robust planning of complex climbing strategies and movements, including extreme movements such as jumping. Similar to recent previous work, we combine a high‐level graph‐based path planner with low‐level sampling‐based optimization of climbing moves. We contribute through showing that neural network models of move success probability, effortfulness, and control policy can make both the high‐level and low‐level components more efficient and robust. The models can be trained through random simulation practice without any data. The models also eliminate the need for laboriously hand‐tuned heuristics for graph search. As a result, we are able to efficiently synthesize climbing sequences involving dynamic leaps and one‐hand swings, i.e. there are no limits to the movement complexity or the number of limbs allowed to move simultaneously. Our supplemental video also provides some comparisons between our AI climber and a real human climber.     

Gender Differences in Mouse and Cursor Movements

Computer cursor and mouse activities such as moving, pointing, selecting, and dragging are essential parts of everyday interactions. Yet it is unknown how men and women differ in the way they move computer cursors. This study examines gender differences in movements of computer cursors. In one experiment, the authors measured trajectories of computer cursors every 20 ms in a simple choice-reaching task and tested the extent to which movement features related to controlling and targeting diverge between male and female participants. Results showed significant gender differences in cursor motions. Female participants deviated from the straight path toward the target location to a larger degree than did male participants, and female participants showed more backward motions (deviating backward from the target location) than did male participants. Implications for sources of these gender differences, user interface and input device design, and musculoskeletal disorders in women are also discussed.     

Geometry and Control of Human Eye Movements

In this paper, we study the human oculomotor system as a simple mechanical control system. It is a well known physiological fact that all eye movements obey Listing's law, which states that eye orientations form a subset consisting of rotation matrices for which the axes are orthogonal to the normal gaze direction. First, we discuss the geometry of this restricted configuration space (referred to as the Listing space). Then we formulate the system as a simple mechanical control system with a holonomic constraint. We propose a realistic model with musculotendon complexes and address the question of controlling the gaze. As an example, an optimal energy control problem is formulated and numerically solved     

头部动作识别系统的硬件设计

介绍了一种用于残疾人的智能交互系统——头部动作识别系统的主体硬件框架,包括头部运动测量单元的设计,采用了三轴陀螺仪L3G4200D和三轴加速度传感器ADXL345.机械臂的执行器驱动单元设计,该部分采用MOSFET驱动,有电流反馈.机械臂结构设计,该部分采用线性执行器电动推杆,推力1500N,速度在2mm/s至60mm/s之间可控,行程200mm,机械臂底盘支持360度全向旋转,整体水平作用距离达1.2m,垂直作用距离1.0m,腕关节支持360度旋转,夹持机构开合行程30mm,可以应对一般家庭应用.为机械臂控制算法的工作提供了硬件平台.     

基于多信息源的股价趋势预测

股票价格及趋势预测是金融智能研究的热门话题。一直以来,各种各样的信息源被不断尝试用于股价预测,例如基本经济特征、技术指标、网络舆情、财务公告、财政新闻、金融研报等。然而,此类研究大多数只使用一种或两种信息源,使用3种及以上信息源的极为少见。信息源越多意味着能够提供更加丰富的信息内容和更多不同的信息层面。但是由于各种信源的本质不同,其对股票市场的影响程度不同,因此将多种信源融合起来进行股价预测并 非易事。此外,多信源也增加了维度灾难的风险。基于信息融合的目的,尝试同时利用基本经济特征、技术指标、网络舆情3种信息源来进行股价预测。具体做法:先对不同类型的信息源数据进行针对性的处理,使其形成统一的数据集,然后使用SVM分类器建立预测模型。实验结果表明,在选用线性核函数和考虑非交易日数据时,使用这3种信源组合的预测模型的预测效果要比使用单一信源或者两两组合的预测效果好。此外,在收集数据时发现,在非交易日(例如周末或停牌期)虽没有买卖但网络舆情剧增。因此,在实验数据中添加了非交易日的舆情情感数据,分类精准度有所提高。研究结果表明,基于多信源融合的股价预测虽然困难,但是在适当地选择特征和针对性地进行数据预处理后会有较好的预测效果。     

跳水动作的视频合成与比较

如何利用计算机视频处理技术,提高体育运动员的科学训练水平,是当前信息处理技术和体育科学技术交叉研究领域的重点问题之一。借助多段运动视频的合成,对各种动作进行观察比较,找出运动员在同一动作上的差异,对于诸如跳水等多项运动而言具有重要的意义。文章根据跳水视频的特点,提出了一种基于运动目标的视频合成算法。通过对不同的方法进行比较实验,结果表明:所提出的算法能很好地完成视频比较,对背景变化具有较强的鲁棒性。另外,算法还可以扩展到多个运动目标比较等应用领域。     

Neural Control of the Movements of a Wheelchair

This paper studies the problem of controlling the movements of a handicapped person's motorized wheelchair from a practical point of view. The control system implemented has been divided into two levels: the low level, consisting of an electronic system which directly controls the drivers of the chair's motors, with a classic PID (proportional-integral-derivative) control loop. The aim of this level is to ensure that the speeds of each one of the wheels is similar to the input speed of these control boards. The second control level (high level), implemented by means of neural techniques, ensures that the linear and angular speeds of the wheelchair are those indicated by a trajectory generator. A new recurrent model is used as the neural network, for which the stability conditions of the complete control system are obtained and various practical tests are carried out, which show the correct performance of the actual system implemented.     

基于六自由度运动测量的下颌运动虚拟再现

为了真实再现下颌的运动过程,设计了一种记录下颌骨六自由度位移的方法;该方法使用两个重叠的网格屏和两束成一定夹角的激光束,用一个CCD视频摄像机采集测量图像,通过简单的计算得到下颌骨的位置与姿态的变化过程;结合上、下牙列的三维扫描数据,可以三维图形的形式再现咬合运动;实验结果表明该文介绍的方法计算公式简单,实时性强,连续测量无误差积累,测量精度可满足口腔医学要求,可用于下颌运动过程的记录。     

一种人眼运动及表情仿真方法

真实感人脸动画是计算机图形学领域的一个研究热点,为了更逼真的仿真虚拟人眼部的表情动作,提出了一种基于机构学原理与肌肉模型的人眼运动及表情仿真方法。可将人眼眼皮的三维网格模型分为边界点和非边界点分别控制,对于边界点的运动采用机构模型实现,对于非边界点及眼部其余部分采用肌肉模型控制,仿真出不同的人眼运动及表情。同时给出了仿真运动的效果实例。采用上述方法可以得到逼真的人眼表情动作仿真动画,效果良好。     

Biomimetic Pinching Movements of a Musculo-Skeletal Dual-Finger Model

The present paper investigates pinching movements using an index finger and a thumb actuated by redundant nonlinear digitorum muscles mimicking the configuration of human fingers. A dual-finger model with 2-d.o.f. joints for each finger and redundant nonlinear digitorum muscles is formulated to mimic the structure of human fingers. First, the kinematics and dynamics of the overall finger–object system, as well as the nonlinear muscular dynamics, are derived in accordance with the results of physiological studies. Next, a sensory-motor control law is proposed to enable stable pinching simultaneously with orientation regulation of an object. This control law includes an internal force term generated by co-construction of the redundant muscles. It is shown that the internal force term can modulate the damping factor in the joint space by its nonlinearity. Based on this effect, it is then shown by numerical simulation that our sensory-motor control law with co-contraction of each digitorum muscle makes it possible to realize pinching movements. Therefore, the pinching movements may be realized by means of a musculo-skeletal dual-finger system with the sensory-motor control law and co-contraction of redundant digitorum muscles.     

复杂运动情况下的多目标检测与追踪

针对多运动目标视频序列中目标存在过缓运动且相互间遮挡的问题,提出一种改进的混合高斯模型(Gaussian Mixture Model,GMM)进行目标检测并结合卡尔曼滤波进行跟踪的新算法;在运动背景检测中,采用三帧差法和混合高斯模型的融合算法进行目标提取,解决了过缓运动目标检测区域不完整的问题;在运动目标跟踪中,由于Kalman滤波器采用目标检测结果进行预测,对观测噪声矩阵进行自适应更新,使得跟踪的稳定性得到加强;通过对比,验证了新算法对存在过缓以及遮挡的不规律运动情况的多目标检测识别与跟踪的实时有效性与准确性。     

发音过程中舌头运动的3D可视化方法*

研究中文发音过程中舌头运动的3D可视化问题。根据核磁共振数据构建舌头精细的3D模型,在此基础上,提取舌背表面处3个点的EMA数据为驱动源,利用弹簧网技术真实再现中文发音过程中的舌头运动。为了验证文中建模和舌头运动合成方法的有效性,使用计算机图形学的方法模拟舌头运动的细节效果,并对比其与由语言学家亲自拍摄的“普通话发音器官动作特征”的X光影像。实验表明,文中方法实现的3D舌头运动符合真实的舌头运动情况,拥有广泛的应用前景。     

Towards Model-Based Recognition of Human Movements in Image Sequences

The interpretation of the movements of articulated bodies in image sequences is one of the most challenging problems in computer vision. In this contribution, we introduce a model-based approach for the recognition of pedestrians. We represent the human body by a 3D-model consisting of cylinders, whereas for modelling the movement of walking we use data from medical motion studies. The estimation of model parameters in consecutive images is done by applying a Kalman filter. Experimental results are shown for synthetic as well as for real image data.