Changle Zhou

Time Series Prediction Based on Linear Regression and SVR

The application of SVR in the time series prediction is increasingly popular. Because some time series prediction based on SVR wasn t very nice in the efficiency of the forecast, this article presents a new regression based on linear regression and SVR. The new regression separates time series into linear part and nonlinear part, then predicts the two parts respectively, and finally integrates the two parts to forecast. Experiments show that the new regression advances the precision of the forecasting compared to the common SVR.

Real-Time Eye Detection in Video Streams

A fast eye detection scheme for use in video streams rather than still images is presented in this paper. The temporal coherence of sequential frames was used to greatly improve the detection speed. First, the eye detector trained by AdaBoost algorithm is used to obtain the rough eye positions. Then these candidate positions are filtered by geometrical patterns of human eyes. The detected eye regions are then taken as the initial detecting window. After each frame is detected, the detecting window is updated. The experiments focused on video stream to exploit the benefits of our detector. In our experiments the mean detection rate was 92.73% for 320 times 240 resolution test videos, with a speed of 24.98 ms per frame. This speed is faster than previous research; however the detection rate does not dramatically decrease.

An Approach to Syndrome Differentiation in Traditional Chinese Medicine based on Neural Network

Although the traditional knowledge representation based on rules is simple and explicit, it is not effective in the field of syndrome differentiation in traditional Chinese medicine (TCM), which involves many uncertain concepts. To represent uncertain knowledge of syndrome differentiation in TCM, two methods were presented respectively based on certainty factors and certainty intervals. Exploiting these two methods, an approach to syndrome differentiation in TCM was proposed based on neural networks to avoid some limitations of other approaches. The main advantage of the approach is that it may realize uncertain inference of syndrome differentiation in TCM, whereas it doesnt request experts to provide all possible combinations for certainty degrees of symptoms and syndromes. Rather than back propagation (BP) algorithm but its modification was employed to improve the capability of generalization of neural networks. First, the standard feedforward multilayer BP neural network and its modification were introduced. Next, two methods for knowledge representation, respectively based on certainty factors and certainty intervals, were presented. Then, the algorithm was proposed based on neural network for the uncertain inference of syndrome differentiation in TCM. Finally, an example was demonstrated to illustrate the algorithm.

Objective Classification Using Advanced Adaboost Algorithm

Adaboost, a general method for improving the accuracy of any given learning algorithm, is usually used to solve the problem of object detection based on cascade structure. However it has some disadvantage. The paper proposes an advanced Adaboost algorithm for object detection. The algorithm adopts a new method to update weighted parameters of weak classifiers. The weights are affected not only by the error rates, but also by their capacity of positive recognition. It is more adaptive to the object detection by decreasing the false alarm rates in the low false rejection rate terminal. The experiment results show the improvement achieved by the new algorithm.

An Intelligent TCM Diagnostic System Based on Intuitionistic Fuzzy Set

This paper proposes an intelligent traditional Chinese medical diagnostic system based on multi-agent system. In this system we also introduce an intelligent fuzzy diagnostic model based on intuitionistic fuzzy set theory according to the characteristics of traditional Chinese medicine, which realize intelligent diagnosis by modeling medical diagnosis rules via fuzzy relations, finally we propose a new approach for similarity measure between the intuitionistic fuzzy sets of syndromes.

Domain-Specific Information Retrieval Based on Improved Language Model

There are two key ingredients in the general framework of language models used in information retrieval, one is importance weighting, the other is word relationship computing. A series of improvements are made to these ingredients of the general framework of language models which is used in domain-specific information retrieval. First, an EM algorithm is proposed to estimate the importance weights of query terms, and the Bayesian smoothing is used to compute the productive probabilities of important terms. Next, a new algorithm based on Dynamic Bayesian Network is proposed for obtaining the explanation probabilities between terms. Experiment shows that the improved model performs remarkably better for domain-specific information retrieval than some traditional retrieval techniques, and the extended framework has good expansibility.

Human Head Modeling Based on an Improved Generic Model

With the development of computer graphics, virtual reality technology is becoming the focus of the research gradually. 3D face modeling technology, which is an important component of the virtual reality technology, has been used in many fields more and more widely. Comparing with the face modeling, the researches in the entire head modeling are less now. This paper proposed an approach to the construction of 3D human head models using an improved generic face model (Candide3) and some 2D human head images. Some simple human expressional controls are realized by adding the animation units (AUs) to the model.

License plate recognition based on intrinsic image decomposition algorithm

Traditional methods of license plate location have a poor effect if they are applied into poor lighting conditions. This paper presents an intrinsic image decomposition algorithm to solve this problem. We extract out the reflection intrinsic image which is independent of light and then conduct license plate location. In addition, to work out the shortage of low recognition rate of the Chinese OCR, this paper proposes the use of R-SIFT feature matching method to authenticate vehicle. For every detecting car, we extract out its feature points using R-SIFT feature matching method and sent to various monitoring stations. When a vehicle image is shot, we first conduct license plate location, and then extract out R-SIFT feature points, finally match with photos in vehicle registration database. If the matched points exceed a set threshold, the system will ring and warn that it may be a suspect vehicle. Thus, we implement the license plate recognition.

A forecast model based on the BP neural network used in refinery's steel equipment's corrosion

The forecasting of the corrosion of refinerys steel equipments shows great importance in preventing the accident. Considering the numerous factors affecting the corroding of refinerys steel equipments, which are uneasily predictable and with complex relationships, this paper proposed a new technology based on the BP neural network technology used in forecasting of the corrosion of refinerys steel equipments. A new model is also built and implemented in this paper. Finally, the experimental results prove the feasibility of the new model and the forecasted results by this new model fixes well with the sample data set.

Diagnosis in Traditional Chinese Medicine Using Artificial Neural Networks: State-of-the-art and Perspectives

Traditional Chinese medicine (TCM), one of Chinas splendid cultural heritages, is the science dealing with human physiology, pathology, diagnosis, treatment and prevention of diseases. With the development of modern science, people come to consider the way of the modernization of TCM. Recently, many researchers mainly in China attempt to realize diagnosis in TCM based on artificial neural networks (DTCMANN). This paper aims at providing an overview of recent DTCMANN studies in TCM field, and focuses on the introduction and summarization of the existing research work about DTCMANN. A review of five major situations, where DTCMANN approaches are applied, has been presented. For each situation, the DTCMANN approaches employed are outlined, as well as the corresponding results. Current research on DTCMANN shows that it is both feasible and promising, and that it is still nearly apiece of virgin soil. The future research direction of DTCMANN is also pointed out based on a discussion of the existing research work