Shaohai Hu

Real-Time Human Behavior Recognition in Intelligent Environment

In recent years, people pay more and more attention to security. Our system presented in this paper is a real-time application to intelligent environment security. It can recognize simple human behaviors and send out alert message intelligently based on human behavior analysis results. This paper mainly describes the behavior analysis methods used in the system, such as moving object detection, human region classification and eigenspace algorithm to recognize human behaviors. Since people usually change their appearances with different dressing, we process images with skeleton algorithm to reduce the impact of appearances. The skeleton structure image sets with all the postures is used to build general eigenspace. Once the general eigenspace is formed, we can recognize behaviors by projecting an unknown human posture into the eigenspace. In our application, six human behaviors (walking, standing, sitting, squatting, leaning and lying) are used. Experimental results show that our method is efficient to recognize these postures

Detecting and tracking of small moving target under the background of sea level

Target detecting and tracking based on video is one of the significant research field of computer vision, in this paper some research has been done on the small moving target under the background of sea level. In the process of target detecting, the moving targets detecting algorithm combining the method of background subtraction with the method of symmetrical difference is proposed in this paper. In the process of target tracking, using mean shift combining Kalman filter. Kalman filter can predict the possible position on the next frame of the video image, and using the mean shift algorithm to search in this neighboring range, this method is effective on fast moving target. The experiment shows that itpsilas a good algorithm to track the small target in the circumstance of low signal noise rate.

Stable parameter settings for PI router mixing TCP and UDP traffic

Research reveals that the fluid-based model can describe the dynamic behavior for bottleneck networks. The stability of the fluid-model can inflect whether there is a congestion of the network. A fluid-based model combined TCP and UDP under PI mechanism has been presented in this paper. We linearized the fluid-model equation and educe two propositions. Based on the fluid model, we linearized this model and added the feedback control mechanism that is introduced by PI, and educed two propositions. Based on two propositions, we can present design guidelines for choosing network parameters that lead to stable operation of the linear feedback control system. We also performed simulations using Matlab and NS2, and the results show that our analysis is suitable.

Two-Dimensional Hybrid Transform (DCT-DWT) for 2-D Signal Processing

Some 2-D signals may have different noise and frequency interference in different directions (horizontal and vertical, or space and time). It is difficult to extract desired signals from such complicated environment of noise and interference by using classical DWT and DCT algorithms. To solve the problem, a hybrid 2-D transform, discrete cosine transform-discrete wavelet (2-D DCT-DWT), with definitions, properties and algorithms, is developed for the hybrid 2-D signal processing. The hybrid transform can be used to remove the noise or interference in 2-D signals from different direction. Under the transform, the 2-D signals are time-frequency ones, one dimensional components to be frequency ones, while the other dimensional components to be time ones. Simulation illustrates the application

SAR image target extraction based on 2-D leapfrog filtering

This paper reveals that the classical R-D imaging algorithm can not remove the strong noise and extract the objects in R-D imaging image by their range and azimuth matching filters. Though 2-D DCT-DWT or 2-D DFT-DWT can be a 2-D filtering for the issue, while the algorithms need huge computation amount, they can not process SAR image real time. To solve the problem, this paper develops a new 2-D filtering algorithm based on 2-D leapfrog filter, which is good at extracting the SAR image target clearly. The 2-D recursive filter is of few multiplications for SAR image each pixel to realize fast and real-time image filtering. Practical experiments of SAR image processing have shown that the approach and algorithm are correct, effective and pragmatical.

The application of 2-D numerical inversion of Laplace transform

For the 2D continuous space-time (CST) systems are described by linear 2-D partial equations, generally, it is impossible to get the closed solutions of the CST systems. To get the space-time response of CST systems needs the inverse 2-D Laplace transform. However, the 2-D inverse Laplace transform (ILT) does not exist for unstable CST, a theorem is proposed to ensure the 2-D ILT to be obtained for stable CST systems. In this paper, we present the approach to get the solutions of CST systems, we also derive an algorithm of numerical 2-D ILT for the vector spatial-time response analysis of CST systems.

2-D DFT-DWT Application to Multidimensional Signal Processing

Some multidimensional signals may have different probability distribution of noise and frequency interference in different directions (horizontal and vertical, or space and time). It poses the question of how to extract desired signals from such complicated environment of noise and interference. It is difficult to extract by using classical DFT and DWT algorithms alone. To solve the problem, a hybrid 2-D transform, discrete Fourier transform-discrete wavelet transform (2-D DFT-DWT), with definitions, properties and algorithms, is developed for the hybrid 2-D signal processing. The hybrid transform can be used to remove the noise or interference in 2-D signals from different directions. Under the transform, desired results can be obtained to take advantage of merits of both DFT and DWT. Also, simulation illustrates the application

2-D Hybrid Wavelet-Fourier Transformation

Based on 1-D continuous wavelet transformation (CWT) and 1-D discrete Fourier transformation (DFT), this paper develops 2-D wavelet-Fourier transformation, which can be used to analyze 2-D continuous-discrete signals and systems in wavelet-Fourier hybrid domain. The definitions and properties of this hybrid 2-D transformation are given in the paper. Also, the numerical algorithms of 2-D wavelet-Fourier transformation are developed for the hybrid 2-D signal processing

A Pervasive Computing Security System Based on Human Activities Analysis

In recent years, the research on intelligent environment based on pervasive/ubiquitous computing technology has attracted more and more attention, especially in security such as KidsRoom by MIT AI Lab. Considering practical applications of pervasive computing technology, it is necessary to present a pervasive computing system model in the view of engineering. In this paper, we illustrated an intelligent surveillant security system based on a HPC model. This system can detect moving person, recognize his behavior, and send out alerting messages such as SMS when intruder or strange behavior is detected. We design this system based on pervasive computing and propose the implementation methods step by step

Signal pre-processing method suitable for compressive sensing based on frequency modulation

The sparse property of the signal to be processed is very important and directly affects the efficiency of compressive sensing. A signal pre-processing method suitable for compressive sensing is given, which is helpful to effective sensing and accurate reconstruction. Under the condition that the sparse characteristic of the signal is unknown, a frequency modulation pattern is introduced to pre-process the signal to increase the sparse proportion of the signal. Then we choose a difference matrix as the reconstruction matrix, the signal could be reconstructed accurately in the process of sparse reconstruction. Theoretical analysis and experimental results show that the proposed pre-processing method for compressive sensing is very effective and efficient.