Probabilistic model-based silhouette refinement for gait recognition

An algorithm to refine and clean gait silhouette noises generated by imperfect motion detection techniques is developed,and a relatively complete and high quality silhouette is obtained.The silhouettes are sequentially refined in two levels according to two different probabilistic models.The first level is within-sequence refinement.Each silhouette in a particular sequence is refined by an individual model trained by the gait images from current sequence.The second level is between-sequence refinement.All t...     

Polling algorithm based on traffic burst characteristics of broadband wireless access systems

In this paper, we propose a novel polling algorithm to decrease the number of idle slots and collission slots during the polling cycle by evaluating the bandwidth request of the subscribe stations (SSs) in the next polling cycle for broadband wireless access (BWA) systems. We firstly analyze the feature of silent time intervals and deduce the Hurst index which symbolizes the degree of self-similarity and long-relative nature. Then we represent the inactive OFF time intervals by the Pareto model and estimate the shape parameter α by the group measured data. Finally we can evaluate the transmission probability of a silent SS before the next polling cycle. By this algorithm, we can find the optimal transmission opportunities for the base station (BS) to achieve the least collision or void timeslots in order to achieve the largest bandwidth efficiency. The theoretical analysis and simulation results show that this polling algorithm can improve the performance for BWA system.     

Human-imitation recognition algorithm based on multi-character

A multi-character recognition method based on hidden Markov model (HMM) was presented. The method can reduce the calculation load of correlation and improve recognition accuracy compared with single-character recognition in video. The characteristics used for recognizing include the shape character, the color character, the texture character and so on. Even our human being generally uses these characteristics to recognize objects in practice. A recognition experiment of 17 fishes was carried out in the paper. The experimental results demonstrate the high veracity of the multi-character recognition algorithm. Together with the tracking process, it can handle dynamic objects, so the multi-character recognition is more like the human recognition, and has great application value.     

Model-based gait representation via spatial point reconstruction

This paper proposed a novel model-based feature representation method to characterize human walking properties for individual recognition by gait. First, a new spatial point reconstruction approach is proposed to recover the coordinates of 3D points from 2D images by the related coordinate conversion factor (CCF). The images are captured by a monocular camera. Second, the human body is represented by a connected three-stick model. Then the parameters of the body model are recovered by the method of projective geometry using the related CCF. Finally, the gait feature composed of those parameters is defined, and it is proved by experiments that those features can partially avoid the influence of viewing angles between the optical axis of the camera and walking direction of the subject. Foundation item: the National Natural Science Foundation of China (No. 60675024)