Yong Haur Tay

Recognizing human gender in computer vision: a survey

Gender is an important demographic attribute of people. This paper provides a survey of human gender recognition in computer vision. A review of approaches exploiting information from face and whole body (either from a still image or gait sequence) is presented. We highlight the challenges faced and survey the representative methods of these approaches. Based on the results, good performance have been achieved for datasets captured under controlled environments, but there is still much work that can be done to improve the robustness of gender recognition under real-life environments.

Two-Stage License Plate Detection Using Gentle Adaboost and SIFT-SVM

This paper presents a two-stage method to detect license plates in real world images. To do license plate detection (LPD), an initial set of possible license plate character regions are first obtained by the first stage classifier and then passed to the second stage classifier to reject non-character regions. 36 Adaboost classifiers (each trained with one alpha-numerical character, i.e. A..Z, 0..9) serve as the first stage classifier. In the second stage, a support vector machine (SVM) trained on scale-invariant feature transform (SIFT) descriptors obtained from training sub-windows were employed. A recall rate of 0.920792 and precision rate of 0.90185 was obtained.

An offline cursive handwritten word recognition system

This paper describes an offline cursive handwritten word recognition system that combines hidden Markov models (HMM) and neural networks (NN). Using a fast left-right slicing method, we generate a segmentation graph that describes all possible ways to segment a word into letters. The NN computes the observation probabilities for each letter hypothesis in the segmentation graph. Then, the HMM compute the likelihood for each word in the lexicon by summing the probabilities over all possible paths through the graph. We present the preprocessing and the recognition process as well as the training procedure for the NN-HMM hybrid system. Another recognition system based on discrete HMM is also presented for performance comparison. The latter is also used for bootstrapping the NN-HMM hybrid system. Recognition performances of the two recognition systems using two image databases of French isolated words are presented. This paper is one of the first publications using the IRONOFF database, and thus can be used as a reference for future work on this database.

Biometric cryptosystems

Despite fuzzy commitment (FC) is a theoretically sound biometric-key binding scheme, it relies on error correction code (ECC) completely to mitigate biometric intra-user variations. Accordingly, FC suffers from the security-performance tradeoff. That is, the larger key size/higher security always trades with poor key release success rate and vice versa. Additionally, the FC is highly susceptible to a number of security and privacy attacks. Furthermore, the best achievable accuracy performance of FC is constrained by the simple distance metrics such as Hamming distance to measure the dissimilarity of binary biometric features. This implies many efficient matching algorithms are to be abandoned. In this paper, we propose an ECC-free key binding scheme along with cancellable transforms for minutiae-based fingerprint biometrics. Apart from that, the minutiae information is favorably protected by a strong non-invertible cancellable transform, which is crucial to prevent a number of security and privacy attacks. The scheme is not limited to binary biometrics as demanded in FC but instead can be applied to various types of biometric features and hence a more effective matcher can be chosen. Experiments conducted on FVC2002 and FVC2004 show that the accuracy performance is comparable to state-of-the-arts. We further demonstrate that the proposed scheme is robust against several major security and privacy attacks. A new ECC-free biometric key binding scheme and the realization in fingerprint biometrics are proposed.A modified randomized GHE in constructing the cancellable transform is proposed.We performed several security and privacy analysis for the proposed scheme, like privacy attacks ARM and SKI.The proposed scheme can be applied to variety of biometric feature representations, not only binary string and matcher.

A review of iris recognition algorithms

Iris recognition has become a popular research in recent years. Due to its reliability and nearly perfect recognition rates, iris recognition is used in high security areas. Among its applications are border control in airports and harbors, access control in laboratories and factories, identification for Automatic Teller Machines (ATMs) and restricted access to police evidence rooms. This paper provides a review of major iris recognition researches. There are three main stages in iris recognition system: image preprocessing, feature extraction and template matching. A literature review of the most prominent algorithms implemented in each stage is presented.

A Comparative Study for Texture Classification Techniques on Wood Species Recognition Problem

Wood species recognition is a texture classification problem that has yet to be well studied. The textures observed on the cross section surface of the wood samples can be used to identify the species of the wood. In this paper, we tested various texture classification techniques, i.e. grey level co-occurrence matrices (GLCM), Gabor filters, combined GLCM and Gabor filters as well as covariance matrix. The experiments are conducted on 512 × 512 images of the six wood species from the CAIRO wood dataset. The experimental results show that the covariance matrix produced using the feature images generated by the Gabor filters is 85% compared to 78.33% for the raw GLCM, 73.33% for the Gabor filters and 76.67% for the combined GLCM and Gabor filters. The experimental results show that the covariance matrix has the best recognition rate.

Detection of license plate characters in natural scene with MSER and SIFT unigram classifier

We present a license plate detector using a fusion of Maximally Stable Extremal Regions (MSER) and SIFT-based unigram classifier trained with Core Vector Machine (CVM). First, MSER is used to obtain a set of regions. Highly unlikely regions are removed with a simplistic heuristic-based filter. Finally, remaining regions with sufficient positively classified SIFT keypoint are retained as likely license plate regions. To train the unigram classifier, a set of SIFT keypoints are obtained from a small set of ground truth images where the license plates are labeled. The training of the SIFT-based unigram classifier is found to be optimal when a CVM is used. On our testing data set, we got a recall rate of 0.98 and a precision rate of 0.964641. On the Caltech Cars (Rear) data set, a recall rate of 0.904762 and precision rate of 0.837349 is obtained.

Online writer identification using character prototypes distributions

Writer identification is a process which aims to identify the writer of a given handwritten document. Its implementation is needed in applications such as forensic document analysis and document retrieval which involved the use of offline handwritten documents. With the recent advances of technology, the invention of digital pen and paper has extended the field of writer identification to cover online handwritten documents. In this communication, a methodology is proposed to solve the problem of text-independent writer identification using online handwritten documents. The proposed methodology would strive to identify the writer of a given handwritten document regardless of its text contents by comparing his or her handwritings with those stored in a reference database. The output of this process would be a ranked list of the writers whose handwritings are stored in the reference database. The main idea is to use the distance measurement between the distributions of reference patterns defined at the character level. Very few, if any, attempts have been done at this character level. Two sets of handwritten document databases each with 82 online documents contributed by 82 subjects were used in the experiments. The reported result was 95% of Top 1 rate accuracy. Only four writers were identified wrongly, ranked as 2, 4, 5 and 12 choice returned.

An analytical handwritten word recognition system with word-level discriminant training

We describe an analytical handwritten word recognition system combining neural networks (NN) and hidden Markov models (HMM). Using a fast left-right slicing method, we generate a segmentation graph that describes all possible ways to segment a word into characters. The NN computes the observation probabilities for each character hypothesis in the segmentation graph. Then, using concatenated character HMMs, a likelihood is computed for each word in the lexicon by multiplying the observation probabilities over the best path through the graph. The role of the NN is to recognize characters and to reject non-characters. We present our approach to globally train the word recognizer using isolated word images. Using a maximum mutual information (MMI) cost function at the word level, the discriminant training updates the parameters of the NN within a global optimization process based on gradient descent. The recognizer is bootstrapped from a baseline recognition system, which is based on character level training. The recognition performance of the globally trained system is compared to the baseline system.

Abnormal Event Detection in Videos Using Spatiotemporal Autoencoder

We present an efficient method for detecting anomalies in videos. Recent applications of convolutional neural networks have shown promises of convolutional layers for object detection and recognition, especially in images. However, convolutional neural networks are supervised and require labels as learning signals. We propose a spatiotemporal architecture for anomaly detection in videos including crowded scenes. Our architecture includes two main components, one for spatial feature representation, and one for learning the temporal evolution of the spatial features. Experimental results on Avenue, Subway and UCSD benchmarks confirm that the detection accuracy of our method is comparable to state-of-the-art methods at a considerable speed of up to 140 fps.