Guangshun Shi

An Expanded Training Set Based Validation Method to Avoid Overfitting for Neural Network Classifier

The overfitting is a problem of fundamental significance with great implications in the applications of neural network. To avoid overfitting, cross-validation has been proposed. However, in many cases the training set is too small so that cross-validation cannot be applied. Aiming at the problem, a new validation method based on expanded training sets is proposed in this paper. Experimental results show that the generalization ability of neural networks can be greatly improved by the proposed validation method.

Research on Offline Palmprint Image Enhancement

Palmprint identification, a subcategory of biometrics identification, has become a hot research area, and image enhancement is a key problem in offline palmprint identification. Since the physiological characteristics and image quality of palmprints are different from those of fingerprints, existing algorithms on fingerprint image enhancement cannot be directly applied in offline palmprint images. Taking into account the characteristics of palmprint images, an enhancement algorithm specific to offline palmprint images is proposed in this paper. We have performed a series of experiments and provide the enhanced palmprint images in the experiment section. Moreover, we evaluate our algorithm by comparing it with the method only using a low-pass filter to smooth the images under the criteria of GI value. Besides, the running time of each step is given to show the efficiency of the algorithm. The result shows that our algorithm is capable of attaining the objectives of offline palmprint enhancement efficiently.

A SVM-HMM Based Online Classifier for Handwritten Chemical Symbols

This paper presents a novel double-stage classifier for handwritten chemical symbols recognition task. The first stage is rough classification, SVM method is used to distinguish non-ring structure (NRS) and organic ring structure (ORS) symbols, while HMM method is used for fine recognition at second stage. A point-sequence-reordering algorithm is proposed to improve the recognition accuracy of ORS symbols. Our test data set contains 101 chemical symbols, 9090 training samples and 3232 test samples. Finally, we obtained top-1 accuracy of 93.10% and top-3 accuracy of 98.08% based on the test data set.

Data Extraction from Deep Web Pages

In this paper, we propose a novel model to extract data from Deep Web pages. The model has four layers, among which the access schedule, extraction layer and data cleaner are based on the rules of structure, logic and application. In the experiment section, we apply the new model to three intelligent system, scientific paper retrieval, electronic ticket ordering and resume searching. The results show that the proposed method is robust and feasible.

A study of on-line handwritten chemical expressions recognition

In this paper, we study the major modules of on-line handwritten chemical expressions recognition. We propose a novel two-level algorithm to recognize expressions. In the first level, structural information is used to distinguish different parts and recognize substances. Then the algorithm segments expressions fatherly and recognizes isolated symbols. To meet the demand of actual applications, the paper also designs an XML-based system to help users save, modify and search the recognition result. The experiment shows that the presented algorithm is reliable.

Location of Special Areas on Palmprint

Offline palmprint plays an important role in personal recognition, especially in the area of criminal investigation. Location of Special Areas is significant to offline palmprint classification and orientation field. This paper proposes an efficient algorithm to locate the Special Areas on offline palmprint, which can be divided into two stages. The first stage is to locate the coarse-level Special Areas based on the topology of four adjacent blocks. And then a dynamic clustering algorithm is adopted to segment the orientation fields of the coarse-level Special Areas according to well-suited optimality criteria. The singular points can be detected based on the information of segmented orientation field. Based on the singular points, the final Special Areas are located. In experimental section, the final location results are not only displayed from the aspect of human inspection, the quantitative results are also provided to compare our location results with official ones and our former algorithm.

Research on Skeletonization of Palmprint Image

Skeletonization is an important procedure of automated palmprint identification system based on the characteristic of minutiae. Skeleton acquired by traditional thinning algorithms will produce many spurious minutiae which are caused by spurs and misconnections between ridges. The excessive erosion is also a common problem in thinning algorithm. Moreover, unit skeleton cannot be produced by some of these algorithms. In this paper, we introduce a thinning algorithm, which is based on Rotation Invariant Thinning Algorithm introduced by Ahmed and Ward [1] (A-W). The proposed method can solve the above problems. The experimental result shows some skeletons from local regions of palmprint images and running time of the algorithm, which shows that the proposed algorithm can achieve better performance without reducing efficiency.

The asymptotic optimization of pre-edited ANN classifier

The generalization problem of an artificial neural network (ANN) classifier with unlimited size of training sample, namely asymptotic optimization in probability, is discussed in this paper. As an improved ANN network model, the pre-edited ANN classifier shows better practical performance than the standard one. However, it has not been widely applied due to the absence of the related theoretical support. To further promote its application in practice, the asymptotic optimization of the pre-edited ANN classifier is studied in this paper. To help study ANN asymptotic optimization in probability, we gives a review of the previous research works on asymptotic optimization in probability of non-parametric classifier, and grouped the main methods into four classes: two-step method, one-step method, generalization method and hypothesis method. In this paper, we adopt generalization/hypothesis mixed method to prove that pre-edited ANN is asymptotically optimal in probability. Furthermore, a simulation is presented to provide an experimental support for our theoretical work.

Recognition of on-line handwritten chemical expressions

In this paper, we study the major modules of on-line handwritten chemical expressions recognition. We propose a novel algorithm that combines two separate methods to segment expressions, one of which is based on structural information and the other on partial recognition. The algorithm improves the traditional algorithm at the stage of recognition, which consists of a substance recognizer and a character recognizer. To meet the demand of actual applications, the paper also designs a standard feature set to deal with the related issues and presents a flexible process of human-computer interaction to help users modify the recognition result. The experimental results show that the presented algorithm is reliable.

The Asymptotic Optimization of the Pre-edited ANN Classifier

The generalization problem of an ANN classifier with unlimited size of training sample, namely asymptotic optimization in probability, is discussed in this paper. As an improved ANN network model, the pre-edited ANN classifier shows better practical performance than the standard one. However, no related theoretical research has been conducted on it. To provide a theoretical support for the pre-edited ANN classifier, the asymptotic optimization is studied in this paper. Furthermore, a simulation is presented to provide an experimental support for our theoretical work.