In this paper, robust visual tracking scheme is achieved through a novel sparse tracking via collaborative motion and appearance (TCMA). A coarse-to-fine framework with both motion and holistic appearance information is taken into consideration. In coarse search, we employ an optical flow map for the generation of motion particles. A rough estimation of target image patch is obtained using l2-regularized least square method in coarse search stage. In fine search, a novel smooth term is proposed in the cost function to improve the robustness of the tracker. With this smooth term, the object appearance in the previous frame will also affect the calculation of sparse coefficient in the current frame. It allows the tracker involving temporal information between consecutive frames instead of only considering single frame appearance information as in the conventional sparse coding-based tracking algorithms. In order to reserve the original and latest appearance information simultaneously in the template, a quadratic-function-like weight allocation scheme combining with particle contributed histogrammic correlation is developed in the updating stage. Both qualitative and quantitative studies are conducted on a set of challenging image sequences. The superior performance over other state-of-the-art algorithms is verified through the experiment.

In this work, a novel approach using Probabilistic CMOS (PCOMS) technology is used to reduce the energy consumption of a fuzzy PID (proportional-integral-derivative) controller. Energy saving is achieved through designing a probabilistic circuit which deliberately reduces the supply voltage of some less significant bits. The fuzzy PID controller consists of 15 bits with floating point representation. Through numerical simulations and VHDL validation, the fuzzy PID can obtain a satisfactory tradeoff with about 4% deviation while achieving a total energy saving of about 32%. Through error analysis, a fuzzy PID is redesigned to tolerate more of randomness in the control signals, hence obtain a better steady state performance while achieving an energy saving of about 51%.

In this paper, we present an active vision system for human-robot interaction purposes that includes robust face detection, tracking, recognition and facial expression analysis. The system will search for human faces in view, zoom on the face of interest based on the face recognition database, track it and finally analyze the emotion parameters on the face. After detection using Haar-cascade classifiers, the variable parameters of the camera are changed adaptively to track the face of the subject by employing the Camshift algorithm, and to extract the facial features which are used for face recognition and facial expression analysis. Embedded Hidden Markov Model is used for face recognition and nonlinear facial mass-spring model is employed to describe the facial muscle’s tension. The motion signatures are then classified using Multi-layer Perceptrons for facial expression analysis. This system can be used as a comprehensive and robust vision package for a robot to interact with human beings.

In this paper, we propose an interactive system for reconstructing human facial expression. In the system, a nonlinear mass-spring model is employed to simulate twenty two facial musclespsila tensions during facial expressions, and then the elastic forces of these tensions are grouped into a vector which is used as the input for facial expression recognition. The experimental results show that the nonlinear facial mass-spring model coupled with the SVM classifier is effective to recognize the facial expressions. Finally, we introduce our robot that can make artificial facial expressions. Experimental results of facial expression generation demonstrate that our robot can imitate six types of facial expressions.

This paper presents a novel control method for a general class of nonlinear systems using neural networks (NNs). Firstly, under the conditions of the system output and its time derivatives being available for feedback, an adaptive state feedback NN controller is developed. When only the output is measurable, by using a high-gain observer to estimate the derivatives of the system output, an adaptive output feedback NN controller is proposed. The closed-loop system is proven to be semi-globally uniformly ultimately bounded (SGUUB). In addition, if the approximation accuracy of the neural networks is high enough and the observer gain is chosen sufficiently large, an arbitrarily small tracking error can be achieved. Simulation results verify the effectiveness of the newly designed scheme and the theoretical discussions.

In this paper, two new parametric network models of robots are presented based on the systems' own functions. The complete dynamics of robots are given in a ready-to-used format which is constructed by finite dimensional static parametric networks for the inertia matrix and the potential energy (or the gravitational forces). As a result, dynamic models of robots can be automatically generated by software once given the number of degrees of freedom (DOF) and the sequence of the joint types, without knowing other parameters such as the lengths and the twist angles of the links. An existing adaptive controller is used as an example to show that some of the controllers can be easily modified such that adaptive controllers can be automatically generated. It is shown that all the closed-loop signals are bounded and tracking error goes to zero.

The nonlinearity of the simplest fuzzy PI controller with different inference methods has been discussed and it shows that Mamdani's minimum inference method is the most general inference method for the fuzzy PI controller. In addition, a novel tuning method based on gain and phase margins has been proposed to determine the weighting coefficients of the fuzzy PI controllers. With the proposed tuning formula, the nature of the fuzzy PI controller has been analyzed and it shows the desired characteristics in terms of gain and phase margins when controlling linear plants. Numerical simulations are presented to show the validity of the proposed tuning method.

This paper discusses the Delta-rule and training of min-max neural networks by developing a differentiation theory for min-max functions, the functions containing min (wedge) and/or max (V) operations. We first prove that under certain conditions all min-max functions are continuously differentiable almost everywhere in the real number field R and derive the explicit formulas for the differentiation. These results are the basis for developing the Delta-rule for the training of min-max neural networks. The convergence of the new Delta-rule is proved theoretically using the stochastic theory, and is demonstrated with a simulation example.