Chun Hung Li

Tongue image matching using color content

This paper addresses the problem of color image matching in medical diagnosis. The color matching of tongue images in different color spaces with different metrics have been investigated and is reported in this article. Two new metrics namely, sorted metric and probabilistic combined metric, are proposed. Existing distance measurements in coordinate space do not satisfy the reflexivity axiom. That means, they are not the valid metrics. To overcome this limitation, the sorted metric in coordinate space is proposed in coordinate space. To improve the matching performance, a probabilistic combined metric is proposed based on the theory of combining classifier. These metrics are applied for the matching of tongue color images and the results are encouraging.

Regularized color clustering in medical image database

A regularized color clustering algorithm is proposed to solve the color clustering problem in medical image database. By incorporating both measures of cluster separability and cluster compactness, regularized color clustering allows the automatic extraction of significant color groups with varying populations. Experimental results in different color spaces show that the regularized color clustering gives superior results in extracting significant distinct/abnormal color clusters without significant increases in cluster compactness. Furthermore, results of color clustering in different color spaces show that the LUV color space is more suitable for color clustering. Methods for selecting the regularization constants have also been suggested.

Locating mobile stations with statistical directional propagation model

Recently, mobile location estimation is drawing considerable attention in the field of wireless communications. Among different mobile location estimation methods, the one which estimates the location of mobile stations with reference to the wave propagation model is drawing much attention. This approach, in principle, makes use of the most primitive property of wave propagation - signal strength, to perform location estimation. Hence this approach should be able to apply to different kinds of cellular network. We found out that in estimating mobile location with reference to signal strength, the azimuth gain of directional antenna and environmental factors can help to improve the accuracy. In this paper, we report our study of a directional propagation model (DPM) which enhanced the traditional propagation model with these factors. We experimented our model with 3,703 sets of real life data collected from a major mobile phone operator of Hong Kong. Results show that the DPM models have significant improvement over other existing location methods in terms of accuracy and stability.

A Semi-supervised SVM for Manifold Learning

Many classification tasks benefit from integrating manifold learning with semi-supervised learning. By formulating the learning task in a semi-supervised manner, we propose a novel objective function that combines the manifold consistency of whole dataset with the hinge loss of class label prediction. This formulation results in a SVM-alike task operating on the kernel derived from the graph Laplacian, and is capable of capturing the intrinsic manifold structure of the whole dataset and maximizing the margin separating labelled examples. Results on face and handwritten digit recognition tasks show significant performance gain. The performance gain is particularly impressive when only a small training set is available, which is often the true scenario of many real-world problems

A directional propagation model for locating mobile stations within a mobile phone network

Mobile location estimation is becoming a high value added service on cellular phone networks. Among different mobile location estimation methods, the one which estimates the location of Mobile Stations (MSs) with reference to an intrinsics of wireless communication, signal strengths, is able to be applied to different kinds of cellular network, and hence is more general. This approach makes use of a wave propagation model to perform location estimation. Most wave propagation models are non-directional. However, we found out that in estimating mobile location with reference to signal strength, the azimuth gain of directional antenna and environmental factors can help to improve the accuracy. In this paper, we report our study of a Directional Propagation Model (DPM) which enhanced the traditional propagation model with these factors. We experimented our model with 192,177 sets of real life data covering most parts of Hong Kong collected from a major mobile hone operator of Hong Kong. Results show that the DPM models have significant improvement over other existing location methods in terms of accuracy and stability.

Network Based Mobile Station Positioning in Metropolitan Area

Recently, mobile station positioning is drawing considerable attention in the field of wireless communications. Many location estimation techniques have been proposed. Although location estimation algorithms based on received signal strength technique may not be the most promising approach for providing location services, signal strength is the only common attribute available among various kind of mobile network. In this paper, we report our study of a Crude Estimation Method (CEM) which estimate the location of a mobile station based on the ratio of the signal strengths received from different base transceiver stations. We conducted series of field tests with the networks of two major mobile phone operators of Hong Kong. Among 6120 real world readings obtained from these field tests, the average errors of CEM is 49.03 meter with a variance of 538.14. When comparing the results of CEM with the Center of Gravity Method (CGM), another mobile location estimation also based on signal strength ratio, CEM has an improvement of 33.85% with about the same variance. These results are very encouraging in the study of mobile network based mobile station positioning in metropolitan areas.

Transductive Learning: Learning Iris Data with Two Labeled Data

This paper presents two graph-based algorithms for solving the transductive learning problem.Sto chastic contraction algorithms with similarity based sampling and normalized similarity based sampling are introduced.The transductive learning on a classical problem of plant iris classification achieves an accuracy of 96% with only 2 labeled data while previous research has often used 100 training samples.The quality of the algorithm is also empirically evaluated on a synthetic clustering problem and on the iris plant data.

Classification of online discussions via content and participation

Web forums and online communities are becoming increasingly important information sources. While there are significant research works on classification of Web pages, relatively less is known on classification of online discussions. By observing the special nature of user participation in web communities, we propose classification methods based on user participation and text content of online discussions. Support vector machines have been employed in this study to classify and analyze discussions based on content and participation. It is found that the accuracy of using participation as classification features can be very high in certain communities. The use of high-dimensional classifier can be effective in enhancing retrieval and classification of online discussion topics.

Spectral Energy Minimization for Semi-supervised Learning

Data mining problems often involve a large amount of unlabeled data and there is often very limited known information on the dataset. In such scenario, semi-supervised learning can often improve classification performance by utilizing unlabeled data for learning. In this paper, we proposed a novel approach to semi-supervised learning as as an optimization of both the classification energy and cluster compactness energy in the unlabeled dataset. The resulting integer programming problem is relaxed by a semi-definite relaxation where efficient solution can be obtained. Furthermore, the spectral graph methods provide improved energy minimization via the incorporation of additional criteria. Results on UCI datasets show promising results.

Semi-supervised Learning in Medical Image Database

This paper presents a novel graph-based algorithm for solving the semi-supervised learning problem. The graph-based algorithm makes use of the recent advances in stochastic graph sampling technqiue and a modeling of the labeling consistency in semi-supervised learning. The quality of the algorithm is empirically evaluated on a synthetic clustering problem. The semi-supervised clustering is also applied to the problem of symptoms classification in medical image database and shows promising results.