Abdesselam Bouzerdoum

A novel skin color model in YCbCr color space and its application to human face detection

This paper presents a new human skin color model in YCbCr color space and its application to human face detection. Skin colors are modeled by a set of three Gaussian clusters, each of which is characterized by a centroid and a covariance matrix. The centroids and covariance matrices are estimated from large set of training samples after a k-means clustering process. Pixels in a color input image can be classified into skin or non-skin based on the Mahalanobis distances to the three clusters. Efficient post-processing techniques namely noise removal, shape criteria, elliptic curve fitting and face/non-face classification are proposed in order to further refine skin segmentation results for the purpose of face detection.

An insect vision-based motion detection chip

The architectural and circuit design aspects of a mixed analog/digital very large scale integration (VLSI) motion detection chip based on models of the insect visual system are described. The chip comprises two one-dimensional 64-cell arrays as well as front-end analog circuitry for early visual processing and digital control circuits. Each analog processing cell comprises a photodetector, circuits for spatial averaging and multiplicative noise cancellation, differentiation, and thresholding. The operation and configuration of the analog cells is controlled by digital circuits, thus implementing a reconfigurable architecture which facilitates the evaluation of several newly designed analog circuits. The chip has been designed and fabricated in a 1.2-/spl mu/m CMOS process and occupies an area of 2/spl times/2 mm/sup 2/.

Adaptive skin segmentation in color images

A new skin segmentation technique for color images is proposed. The proposed technique uses a human skin color model that is based on the Bayesian decision theory and developed using a large training set of skin colors and nonskin colors. The proposed technique is novel and unique in that texture characteristics of the human skin are used to select appropriate skin color thresholds for skin segmentation. Two homogeneity measures for skin regions that take into account both global and local image features are also proposed. Experimental results showed that the proposed technique can achieve good skin segmentation performance (false detection rate of 4.5% and false rejection rate of 4.0%).

A generalized feedforward neural network architecture for classification and regression

This article presents a new generalized feedforward neural network (GFNN) architecture for pattern classification and regression. The GFNN architecture uses as the basic computing unit a generalized shunting neuron (GSN) model, which includes as special cases the perceptron and the shunting inhibitory neuron. GSNs are capable of forming complex, nonlinear decision boundaries. This allows the GFNN architecture to easily learn some complex pattern classification problems. In this article the GFNNs are applied to several benchmark classification problems, and their performance is compared to the performances of SIANNs and multilayer perceptrons. Experimental results show that a single GSN can outperform both the SIANN and MLP networks.

A universal and robust human skin color model using neural networks

We propose a new image classification technique that utilizes neural networks to classify skin and non-skin pixels in color images. The aim is to develop a universal and robust model of the human skin color that caters for all human races. The ability to detecting solid skin regions in color images by the model is extremely useful in applications such as face detection and recognition, and human gesture analysis. Experimental results show that the neural network classifiers can consistently achieve up to 90% accuracy in skin color detection.

Skin segmentation using color and edge information

An algorithm for segmenting skin regions in color images using color and edge information is presented. Skin colored regions are first detected using a Bayesian model of the human skin color. These regions are further segmented into skin region candidates that satisfy the homogeneity property of the human skin. We show that Bayesian skin color model outperforms many other models such as the piece-wise linear models, Gaussian models and model based on multilayer perceptrons. Experimental results indicate that the proposed segmentation algorithm reduces false detection caused by background pixels having skin colors, and more significantly it is capable of separating true skin regions from falsely detected regions.

Classification of digital modulation schemes using neural networks

Modulation recognition systems have to be able to correctly classify the incoming signals modulation scheme in the presence of noise. This paper addresses the problem of automatic modulation recognition of digital communication signals using neural networks. Seven digital modulation schemes have been considered and seven features have been used as inputs to the neural network (NN) to perform the classification. Several NN structures have been tested that perform at over 99% accuracy at signal-to-noise ratios (SNR) of 10 dB. Design considerations for the NN classifier are discussed and the implementation of these has been shown to produce significant reduction in network size. The performance of the NN-based classifier has also been compared with that of a decision-theoretic classifier; it was found that the NN slightly outperforms the decision-theoretic classifier.

Application of shunting inhibitory artificial neural networks to medical diagnosis

Shunting inhibitory artificial neural networks (SIANNs) are biologically inspired networks in which the neurons interact among each other via a nonlinear mechanism called shunting inhibition. Since they are high-order networks, SIANNs are capable of producing complex, nonlinear decision boundaries. In this article, feedforward SIANNs are applied to several medical diagnosis problems and the results are compared with those obtained using multilayer perceptrons (MLPs). First, the structure of feedforward SIANNs is presented. Then, these networks are applied to some standard medical classification problems, namely the Pima Indians diabetes and Wisconsin breast cancer classification problems. The SIANN performance compares favourably with that of MLPs. Moreover, some problems with the diabetes data set are addressed and a reduction in the number of inputs is investigated.

Classification and function approximation using feed-forward shunting inhibitory artificial neural networks

In this article we propose a new class of artificial neural networks for classification and function approximation. These networks are referred to as shunting inhibitory artificial neural networks (SIANN). A SIANN consists of one or more hidden layers comprised of shunting neurons, the outputs of which are combined linearly to form the desired output. The basic synaptic interaction of the hidden units is shunting inhibition. Due to the inherent nonlinearity mediated by shunting inhibition, SIANN networks are capable of constructing a large repertoire of decision surfaces, ranging from simple hyperplanes to very complex nonlinear hypersurfaces. Therefore, developing efficient training algorithms for these networks should simplify the design of very powerful classifiers and function approximators. In this paper some examples of complex decision regions formed by SIANN are illustrated. Furthermore, a method for training feedforward SIANN is developed based on the error backpropagation algorithm. Finally, simulation results which illustrate the performance of SIANN in function approximation and classification tasks are presented and compared with results obtained from multilayer perceptron networks.

A face detection system using shunting inhibitory convolutional neural networks

We present a face detection system based on a class of convolutional neural networks, namely shunting inhibitory convolutional neural networks (SICoNNets). The topology of these networks is a flexible feedforward architecture with three different connections schemes: fully-connected, toeplitz-connected and binary-connected. SICoNNets were trained, using a hybrid method based on Rprop, Quickprop and least squares, to discriminate between face and non-face patterns. All three connection schemes achieve 99% detection accuracy at 5% false alarm rate, based on a test set of 7000 face and non-face patterns. Furthermore, toeplitz-connected network was trained on a larger training set and has achieved a 99% correct classification rate with only 1% false alarm rate based on the same test set. A face detection system is built based on the trained convolutional neural networks. The system accepts an input image of arbitrary size and localizes the face patterns in the image. To localize faces of different sizes, the convolutional neural network is applied as a face detection filter at different scales. The detection scores from different scales are aggregated together to form the final decision.