Zhenjiang Miao

Projection Histogram Based Human Posture Recognition

As different posture has different projection histogram pattern, the projection histogram can be used as one of the features to discriminate different postures. In this paper, a new method using projection histogram for static human posture recognition is proposed. It comprises of three key modules: background subtraction, projection histogram computing and template matching. Comparing with many other methods, our approach is fast, simple and less sensitive to noise. Using our new method, a system is implemented and tested with ten static postures. It can automatically recognize them with high percentage of right decisions

Real Time Multiple Object Tracking Using Tracking Matrix

Multi-object tracking is an important subject and challenging task in computer vision research. This article presented a method of multiple objects tracking for real-time intelligent surveillance. After object detection, using Kalman filter to predict objects’ state. Then, a “tracking matrix” is calculated based on color histogram information to establish the corresponding relationship between objects. When there is occlusion or splitting, “mother object” and “child object” are introduced to maintain continuous and reliable tracking. Experiment results show that the proposed methods are fast and effective.

Does linear combination outperform the k-NN rule?

Some classifier combination experimental results show that the classification error rate of one linear combination method, namely multi-response linear regression is smaller than that of classical k-NN rule. This paper discusses the reason which results in this phenomenon and proposes a new training data set edit approach to improve the performance of the k-NN rule. Our new approach is tested on two large data sets selected from ELENA database and UCI database, the experimental results show it outperform both classical k-NN and linear regression

Research on Vehicle Type Classification Based on Spatial Pyramid Representation and BP Neural Network

This paper presents a method of the vehicle type classification based on spatial pyramid representation and BP neural network. We extract feature vectors of each vehicle image by using the spatial pyramid representation method. By this way, we can use different size of pictures instead of changing the picture into a fixed size avoiding the deformation of the target images when cropping or warping and so on. We choose BP neural network to train our classifier and have a good performance on car, bus and truck classification.

PDE-Based Finger Image Denoising

Finger image has attracted the interesting of researchers in biometrics recently, and the lines in it can be used for person recognition. In order to detect lines from the finger image, this paper proposes a new image denoising method. Firstly, the features of the lines in finger image are studied, and an effective edge detection algorithm is put forward. Then, the partial differentiation equations (PDE) for image denoising are defined based on the edge detection algorithm. Finally, the finite differences scheme for the PDEs is given and several experimental results are presented.

Error correction via phonetic similarity-based processing for chinese spoken dialogue system

This paper reviewed the approaches of error handling in spoken dialogue systems in the literature at first. To take advantage of domain knowledge and the characteristics of the errors in Chinese speech recognition, a similarity between text strings from speech recognizer and the correct words was defined. A similarity-based algorithm of error handling in Chinese spoken language understanding was developed to improve the correctness and robustness of sentence understanding in spoken dialog systems. Before the procedure of keyword spotting, the errors in text strings are revised first. Preliminary experimental results show that the algorithm is effective

A fast Genetic Fuzzy approach and its application in Classifier Fusion

This paper proposes a new genetic fuzzy approach, in which the traditional method for generating the initial population and the cross operator are improved. Our new method is tested on classifier fusion problem, and the experimental results show it outperform other classifier combination methods in the classification accuracy

High speed markerless monocular expression cloning

Expression cloning is essential to motion capture. Currently, most methods to realize the idea require expensive equipment such as Cyberware and lot of manual work. This paper introduces a high speed monocular markerless expression cloning system. In Motion capture module, feature points in actors face are captured and tracked with computer vision methods. Tracking result of 16 feature points is encouraging after only a few steps of manual work. In the Expression driven model, we well design a template model, then additional models can be loaded without initialization. Only a camera and a common PC is needed in our system, on condition that source video is taken under normal lightning environment, synchronous realistic target animation can be acquired.

Design and Implementation of a Vision-Based Motion Capture System

Motion Capture has been widely used in many fields such as animation and game production. This paper describes the scheme and implementation of a self-designed vision-based motion capture system. This system is set up based on a Client/Server network structure. The client software tracks every marker attached to the actor’s body and sends the results to the server in real-time. The server software recovers every joint’s 3D position and shows the capturing results simultaneously. It can also convert the 3D position data into BVH file data.

Multi-modal human state perception for pervasive computing

Pervasive/ubiquitous computing is anywhere and anytime human-centered services based on humans needs accordingly. Multi-modal human state perception for necessary context awareness is a key issue to build such a human-centered pervasive computing environment. In this paper, we design and implement a platform for human visual and physiological perception which gives human identity facial expression and physiological information directly. By integrating and modeling the face and physiological state information, we can also have some possible deep emotion information. These contexts are used for further pervasive computing applications such as healthcare