Chen-Chiung Hsieh

Off-line recognition of handwritten Chinese characters by on-line model-guided matching

Abstract A model-guided structural matching method for recognizing handwritten off-line Chinese characters is presented. According to the stroke writing sequence, each character is described by an on-line model, which is a one-dimensional (1D) string consisting of a stroke sequence interleaved with relationships between two consecutive strokes. After an unknown input is thinned and line approximated, all possible strokes are extracted. A tree is created to represent all possible combinations of the extracted strokes which have the same stroke sequence as the one defined in the model. In the recognition process, we match the strokes with those defined in the on-line model. The matching process is formulated as a tree searching algorithm guided by the relationships in the model. The modelled character is taken to be a candidate if there is a feasible path which satisfies the relationships defined in the on-line model and the number of missing strokes in the input is less than a given threshold. The input is recognized as the character with the greatest number of matched strokes among all candidates. Experimental results on 300 frequently used characters in a database called CCL/HCCR1, which contains 5401 Chinese characters and about 250 variations for each one, show that the recognition rate is over 90%.

Fast enhanced face-based adaptive skin color model

Man machine interface by video analysis becomes popular recently. The most typical body gesture utilized for computer interaction is hand gesture. Therefore, it is a very important topic to accurately extract hand regions from a sequence of images in real time. In this paper, we propose an adaptive skin color model which is based on detected face color. Skin colors are sampled from extracted face region where non-skin color pixels like eyebrow or glasses are excluded. Gaussian distributions of normalized RGB are then used to define the skin color model for the detected person. To demonstrate the robustness of proposed model, experiments under diversified lighting and background are tested. Traditional methods based on RGB, Normalized RGB, and YCbCr are all implemented for comparison. From experimental results, skin color pixels could be detected for each person. The accuracy rate is 95.73% on average and is superior to previously mentioned methods.

A PROBABILISTIC STROKE-BASED VITERBI ALGORITHM FOR HANDWRITTEN CHINESE CHARACTERS RECOGNITION

This paper proposes a probabilistic approach to recognize handwritten Chinese characters. According to the stroke writing sequence, strokes and interleaved stroke relations are built manually as a 1-D string, called an on-line model, to describe a Chinese character. In an input character, strokes are first extracted by a tree searching method. The recognition problem is then formulated as an optimization matching problem in a multistage directed graph, where the number of stages is the length of the modelled stroke sequence. Nodes in a stage represent extracted strokes that have the same stroke type as defined in the on-line model and the link between two neighboring nodes corresponds to the relationship between the two extracted strokes. The probability that the extracted stroke belongs to the predefined stroke type is calculated from the stroke line segments, and the transition probability between two extracted strokes is the degree of satisfaction of the relationship defined in the on-line model. The Viterbi algorithm, which can handle stroke insertion, deletion, splitting, and merging, is applied to recover the sequence of strokes consisting of the unknown character. The similarity is defined to be the product of stroke probabilities and stroke transition probabilities in the stroke sequence. The unknown character is matched with all modelled characters and is recognized as the one with the highest similarity. Experiments with 540 characters uniformly selected from the database CCL/HCCR1 (250 variations/class) are conducted, and the recognition rate is about 92.8%, which proves the feasibility of the proposed recognition system.

A Facial Expression Classification System Based on Active Shape Model and Support Vector Machine

Most traditional expression classification systems track facial component regions such as eyes, eyebrows, and mouth for feature extraction. This paper utilized facial components to locate dynamic facial textures such as frown lines, nose wrinkle patterns, and nasolabial folds to classify facial expressions. Adaboost using Haar-like feature and Active Shape Model (ASM) are adopted to accurately detect face and acquire important facial feature regions. Gabor filter and Laplacian of Gaussian are used to extract texture information in the acquired feature regions. These texture feature vectors represent the changes of facial texture from one expression to another expression. Support Vector Machine is deployed to classify the six facial expression types including neutral, happiness, surprise, anger, disgust, and fear. Cohn-Kanade database was used to test the feasibility of proposed method and the average recognition rate reached 91.7%.

A reinforcement-learning approach to robot navigation

This paper presents a reinforcement-learning approach to a navigation system which allows a goal-directed mobile robot to incrementally adapt to an unknown environment. Fuzzy rules which map current sensory inputs to appropriate actions are built through the reinforcement learning. Simulation results illustrate the performance of the proposed navigation system. In this paper, ACSNFIS is used as the main network architecture to implement the reinforcement-learning based navigation system.

Integrating virtual objects into real images for augmented reality

1. ABSTRACT A systematic approach for integrating virtual objects into real images is developed in this paper. We propose the P3P-IC. method to solve the camera pose estimation problem. A robust tracking method is developed via the combination of the LMedS technique and the P3P-ICP method. With the 3D models and the robust tracking methods, we can determine the camera poses associated with each frame in the image sequence. Knowing the camera poses for each image frame, we can then integrate virtual objects into a video segment. 1.1 Keywords Augmented reality, computer graphics, computer vision.

A Simple and Fast Surveillance System for Human Tracking and Behavior Analysis

In this paper, we designed a simple and fast visual surveillance system to track human position and to determine if any abnormal behavior like wall climbing and falling happened. By taking both time and background difference into considerations, illumination effects could be greatly reduced while calculating motion masks. Refinements including holes filling, shadow removal, and noise reduction are done to obtain much more reliable motion masks. However, motion masks corresponding to occluded moving people, greater than a given width, are segmented recursively into smaller ones by bi-modal thresholding. Meanwhile, background could also be updated by the refined motion masks. Integrated location-based and weighted block-based matching is done for object tracking. A similarity is defined from these weighted matched block for object classification. Finally, a couple of criterions are defined to analyze whether objects stop, disappear, climb, or fall. Experimental results are given to demonstrate the robustness of our system.

Anti-SIFT Images Based CAPTCHA Using Versatile Characters

Due to vigorous development of pattern recognition, traditional human form filling tasks would be replaced by automated processes. However, these automation processes are often misused for illegal behavior such as spam e-mail or application for website account. In order to prevent website owner from suffering the attacks of automated program, this paper proposed an innovative image-based CAPTCHA for distinguishing human and computer by embedding versatile characters in the images. The proposed method makes the characters indiscernible by automated image analysis technologies like scale-invariant feature transform while human can easily distinguish the location of the embedded characters. Our designed mechanism was capable to elude such kind of attacks. In experiments, 15 users were invited to test the system and the success rate is 86%. If wrong operations like clicking out of text boxes were excluded, the success rate reached 95%. Compare the average logging time with reCAPTCHA and HELLO CAPTCHA, the proposed method is faster than these two methods by 32 seconds and 115 seconds, respectively.

Design and implementation of the interactive multimedia broadcasting services in DVB-H

DVB-H is a mobile digital video broadcasting technology based on DVB-T with the advantages of time-slicing and multiprotocol encapsulation¿forward error correction (MPE-FEC) to reduce the power consumption and enhance the stability of receiver. The encapsulated data is compatible with the Internet Protocol which makes the integration of broadcast and telecommunication much easier. Therefore, digital content can be transmitted by adding encapsulation for each layer. In this paper, we developed a portable middleware for DVB-H clients. It could parse the Transport Stream according to the IP Data Casting (IPDC) protocol stack to achieve interactive multimedia services. Moreover, channel information contained in Electronic Service Guide (ESG) could also be decoded. Several applications were developed on our middleware to demonstrate its feasibility as an interactive multimedia platform.

A Reverse-Order Scheduling Scheme for Broadcasting Continuous Multimedia Data Over a Single Channel

In this work, we propose a reverse-order scheduling (ROS) method for broadcasting a video over a single channel. The method firstly partitions the video into equal-length segments, then divides these segments into groups, and finally broadcasts the segments in the same group over the same subchannel in the reverse order of their indices. Further, we also provide a thorough analysis of the performance of the ROS method. The main contribution of the ROS method is to exhibit not only a shorter waiting time but also a smaller buffer requirement. Compared with the Alternative Broadcasting (AB), Hopping-Insertion (HI), and SingBroad methods, the proposed ROS method reduces the average waiting time by 38%~96%, 18%~63%, and 3%~18%, respectively, while the bandwidth of a single channel is capable of more than triple the consumption rate. Further, the ROS method reduces the buffer requirement by 30%~70% and 30%~55% as compared with the AB and SingBroad methods, respectively, while the buffer requirement of the HI method is not given.