Jeffrey Huang

Integrating independent components and linear discriminant analysis for gender classification

Computer vision and pattern recognition systems play an important role in our lives by means of automated face detection, face and gesture recognition, and estimation of gender and age. We have developed a gender classifier with performance superior to existing gender classifiers. This paper addresses the problem of gender classification using frontal facial images. The testbed consists of 500 images (250 females and 250 males) randomly withdrawn from the FERET facial database. Independent component analysis (ICA) is used to represent each image as a feature vector in a low dimensional subspace. A classifier based on linear discriminant analysis (LDA) is used in this lower dimensional subspace. Our experimental results show a significant improvement in gender classification accuracy and we obtain an accuracy of 99.3%.

Visual routines for eye location using learning and evolution

Eye location is used as a test bed for developing navigation routines implemented as visual routines within the framework of adaptive behavior-based AI. The adaptive eye location approach seeks first where salient objects are, and then what their identity is. Specifically, eye location involves: 1) the derivation of the saliency attention map, and 2) the possible classification of salient locations as eve regions. The saliency (where) map is derived using a consensus between navigation routines encoded as finite-state automata exploring the facial landscape and evolved using genetic algorithms (GAs). The classification (what) stage is concerned with the optimal selection of features, and the derivation of decision trees, using GAs, to possibly classify salient locations as eyes. The experimental results, using facial image data, show the feasibility of our method, and suggest a novel approach for the adaptive development of task-driven active perception and navigational mechanisms.

Gender identification using frontal facial images

Computer vision and pattern recognition systems play an important role in our lives by means of automated face detection, face and gesture recognition, and estimation of gender and age. This paper addresses the problem of gender classification using frontal facial images. We have developed gender classifiers with performance superior to existing gender classifiers. We experiment on 500 images (250 females and 250 males) randomly withdrawn from the FERET facial database. Independent component analysis (ICA) is used to represent each image as a feature vector in a low dimensional subspace. Different classifiers are studied in this lower dimensional space. Our experimental results show the superior performance of our approach to the existing gender classifiers. We get a 96% accuracy using support vector machine (SVM) in ICA space.

Automated diagnosis of fetal alcohol syndrome using 3D facial image analysis

OBJECTIVESnUse three-dimensional (3D) facial laser scanned images from children with fetal alcohol syndrome (FAS) and controls to develop an automated diagnosis technique that can reliably and accurately identify individuals prenatally exposed to alcohol.nnnMETHODSnA detailed dysmorphology evaluation, history of prenatal alcohol exposure, and 3D facial laser scans were obtained from 149 individuals (86 FAS; 63 Control) recruited from two study sites (Cape Town, South Africa and Helsinki, Finland). Computer graphics, machine learning, and pattern recognition techniques were used to automatically identify a set of facial features that best discriminated individuals with FAS from controls in each sample.nnnRESULTSnAn automated feature detection and analysis technique was developed and applied to the two study populations. A unique set of facial regions and features were identified for each population that accurately discriminated FAS and control faces without any human intervention.nnnCONCLUSIONnOur results demonstrate that computer algorithms can be used to automatically detect facial features that can discriminate FAS and control faces.

Integrating independent components and support vector machines for gender classification

Computer vision and pattern recognition systems play an important role in our lives by means of automated face detection, face and gesture recognition, and estimation of gender and age. We have developed a gender classifier with performance superior to existing gender classifiers. This paper addresses the problem of gender classification using frontal facial images. The testbed consists of 500 images (250 females and 250 males) randomly withdrawn from the FERET facial database. Independent component analysis (ICA) is used to represent each image as a feature vector in a low dimensional subspace. Different classifiers are studied in this lower dimensional subspace. Our experimental results show the best accuracy of 96% in gender classification by combining ICA and support vector machines (SVMs).

A comparative study of genetic sequence classification algorithms

Classification of genetic sequence data available in public and private databases is an important problem in using, understanding, retrieving, filtering and correlating such large volumes of information. Although a significant amount of research effort is being spent internationally on this problem, very few studies exist that compare different classification approaches in terms of an objective and quantitative classification performance criterion. In this paper, we present experimental studies for classification of genetic sequences using both unsupervised and supervised approaches, focusing on both computational effort as well as a suitably defined classification performance measure. The results indicate that both unsupervised classification using the Maximin algorithm combined with FASTA sequence alignment algorithm and supervised classification using artificial neural network have good classification performance, with the unsupervised classification performs better and the supervised classification performs faster. A trade-off between the quality of classification and the computational efforts exists. The utilization of these classifiers for retrieval, filtering and correlation of genetic information as well as prediction of functions and structures will be logical future directions for further research.

Using facial images to diagnose fetal alcohol syndrome (FAS)

This paper proposes the methodology of classification architectures for FAS diagnosis tasks and shows their feasibility through experimental studies. We describe the automatic selection of features from an image training set using the theories of multidimensional discriminant analysis and the associated optimal linear projection. The method consists of two steps: projection of face image from the original vector space to a face subspace via PCA, and then use of LDA to obtain a linear classifier. This hybrid classifier using PCA and LDA provides an effective framework for classification of FAS images as FAS-positive and FAS-negative.

Comparative Performance Evaluation of Gray-Scale and Color Information for Face Recognition Tasks

This paper assesses the usefulness of color information for face recognition tasks. Experimental results using the FERET database show that color information improves performance for detecting and locating eyes and faces, respectively, and that there is no significant difference in recognition accuracy between full color and gray-scale face imagery. Our experiments have also shown that the eigenvectors generated by the red channel lead to improved performance against the eigenvectors generated from all the other monochromatic channels. The probable reason for this observation is that in the near infrared portion of the electro-magnetic spectrum, the face is least sensitive to changes in illumination. As a consequence it seems that the color space as a whole does not improve performance on face recognition but that when one considers monochrome channels on their own the red channel could benefit both learning the eigenspace and serving as input to project on it.

Digital geometry image analysis for medical diagnosis

This paper describes a new medical image analysis technique for polygon mesh surfaces of human faces for a medical diagnosis application. The goal is to explore the natural patterns and 3D facial features to provide diagnostic information for Fetal Alcohol Syndrome (FAS). Our approach is based on a digital geometry analysis framework that applies pattern recognition techniques to digital geometry (polygon mesh) data from 3D laser scanners and other sources. Novel 3D geometric features are extracted and analyzed to determine the most discriminatory features that best represent FAS characteristics. As part of the NIH Consortium for FASD, the techniques developed here are being applied and tested on real patient datasets collected by the NIH Consortium both within and outside the US.

Mining p53 binding sites using profile hidden Markov model

Hidden Markov model has been successfully applied to bacterial gene finders and mRNA splicing modeling. Using a set of observing DNA sequences, HMM is derived for homologous search. In this paper we develop profile HMM in detecting p53, a tumor suppressor, binding sites along genes. Without assuming the constant number of nucleotides in p53 binding site, profile HMM and viterbi algorithms are designed to detect the embedded p53 binding sites from the promoter genes chosen from GenBank. The p53 regulated genes containing either single or multiple p53 binding sites distributed as clusters can be identified and classified into 7 functional groups including cell cycle regulation, DNA damage repair, signaling transduction, transcriptional factor, stress response, tumor suppressor, and oncogen.