Jun Guo

HCL2000 - A Large-scale Handwritten Chinese Character Database for Handwritten Character Recognition

In this paper, we present a large scale off-line handwritten Chinese character database-HCL2000 which will be made public available for the research community. The database contains 3,755 frequently used simplified Chinesecharacters written by 1,000 different subjects. The writers’ information is incorporated in the database to facilitate testing on grouping writers with different background such as age, occupation, gender, and education etc. We investigate some characteristics of writing styles from different groups of writers. We evaluate HCL2000 database using three different algorithms as a baseline. We decide to publish the database along with this paper and make it free for a research purpose.

Web Multimedia Object Classification Using Cross-Domain Correlation Knowledge

Given a collection of web images with the corresponding textual descriptions, in this paper, we propose a novel cross-domain learning method to classify these web multimedia objects by transferring the correlation knowledge among different information sources. Here, the knowledge is extracted from unlabeled objects through unsupervised learning and applied to perform supervised classification tasks. To mine more meaningful correlation knowledge, instead of using commonly used visual words in the traditional bag-of-visual-words (BoW) model, we discover higher level visual components (words and phrases) to incorporate the spatial and semantic information into our image representation model, i.e., bag-of-visual-phrases (BoP). By combining the enriched visual components with the textual words, we calculate the frequently co-occurring pairs among them to construct a cross-domain correlated graph in which the correlation knowledge is mined. After that, we investigate two different strategies to apply such knowledge to enrich the feature space where the supervised classification is performed. By transferring such knowledge, our cross-domain transfer learning method can not only handle large scale web multimedia objects, but also deal with the situation that the textual descriptions of a small portion of web images are missing. Empirical experiments on two different datasets of web multimedia objects are conducted to demonstrate the efficacy and effectiveness of our proposed cross-domain transfer learning method.

Detection of vehicle manufacture logos using contextual information

Besides the decorative purposes, vehicle manufacture logos can provide rich information for vehicle verification and classification in many applications such as security and information retrieval. Detection and recognition of vehicle manufacture logos are, however, very challenging because they might lack of discriminative features themselves. In this paper, we propose a method to detect vehicle manufacture logos using contextual information, i.e., the information of surrounding objects near vehicle manufacture logos such as license plates, headlights, and grilles. The experimental results demonstrate that the proposed method is more effective and robust than other methods.

Web Multimedia Object Clustering via Information Fusion

Multimedia information plays an increasingly important role in humans daily activities. Given a set of web multimedia objects (images with corresponding texts), a challenging problem is how to group these images into several clusters using the available information. Previous researches focus on either adopting individual information, or simply combining image and text information together for clustering. In this paper, we propose a novel approach (Dynamic Weighted Clustering) to separate images under the supervision of text descriptions, Also, we provide a comparative experimental investigation on utilizing text and image information to tackle web image clustering. Empirical experiments on a manually collected web multimedia object (related to the events after disasters) dataset are conducted to demonstrate the efficacy of our proposed method.

A multimedia information fusion framework for web image categorization

With the rapid development of technologies for fast Internet access and the popularization of digital cameras, an enormous number of digital images are posted and shared online everyday. Web images are usually organized by topic and are often assigned appropriate topic-related textual descriptions. Given a large set of images along with the corresponding texts, a challenging problem is how to utilize the available information to efficiently and effectively perform image retrieval tasks, such as image classification and image clustering. Previous approaches on image categorization focus on either adopting text or image features, or simply combining these two types of information together. In this paper, we improve our previously reported two multi-view classification approaches—(Dynamic Weighting and Region-based Semantic Concept Integration) for categorizing the images under the “supervision” of topic-related textual descriptions—by proposing a novel multimedia information fusion framework, in which these two proposed methods are seamlessly integrated by analyzing the special characteristics of different images. Notice that, the proposed framework is a generic multimedia information fusion framework which is not limited to our previously reported two approaches, and it can also be used to integrate other existing multi-view classification methods or models. Also, our proposed framework is capable of handling the large scale image categorization. Specifically, the proposed framework can automatically choose an appropriate classification model for each testing image according to its special characteristics and consequently achieve better classification performance with relatively less computation time for large scale datasets; Moreover, it is able to categorize images without any textual description in real world applications. Empirical experiments on two different types of web image datasets demonstrate the efficacy and efficiency of our proposed classification framework.

Efficient a priori SNR estimation based on parameter adaptive spectral method

Abstract The a priori signal-to-noise (SNR) is one of the most important parameters in the short-time spectrum estimation techniques in speech enhancement. A new and convenient algorithm to estimate the priori SNR is involved in this paper. In this paper, the priori and posterior SNR of intra-frame are defined which can trace the variation of the a priori SNR of each frame better and can solve the problem of delay involved by the traditional approaches. Simulation shows that, the performance of the proposed algorithm is better than the traditional estimators in terms of log-spectral distance and the improvement segmental SNR, especially in the no stationary noise environments.

Burst feature detection using parameter estimated two-state automaton

Abstract In recent text mining research, there is a trend in analyzing the burst features of specific entity such as a word, a meme or a document in text streams. Such burst features can be efficiently and robustly identified by Kleinbergs two-state automaton model. However, the two parameters of the model, which is manually set, have heavily affected the performance of the model. In this paper, the function of the two parameters is examined, and two algorithms are proposed for the estimation of the two parameters. Experiments with public news corpora prove that our estimation can maximize the reliability of the detection results and remove the noisy burst features effectively.

N-LBP based vehicle monitoring system

In recent years, feature based object detection has attracted increasing attention in computer vision research community. However, to our best knowledge, no previous work has focused on utilizing local binary pattern (LBP) for vehicle detection in Intelligent Transportation System(ITS) domain. In this paper, we develop a novel traffic monitoring system based on N-LBP algorithm, which is the new LBP texture descriptor proposed. The approach includes three steps: firstly the general critical ingredients (GCI for short) are selected from LBP features through training to indicate vehicles. Then GCI are extracted from region of interest (ROI) in the new image for object detection and identification. Linear Kalman filter is employed for feature based tracking finally. Experimental results demonstrate the superiority of N-LBP feature over basic LBP feature, and performance of the new system is more stable and reliable.