Face Detection using a Hybrid Neural Network Model

H.-J. Kim and H.-S. Yang (Korea)


Face detection, FMM model, hybrid neural network


In this paper, we present a face detection method using a hybrid neural network model. The method is a three-stage process which consists of preprocessor, feature extractor and pattern classifier. For the preprocessor, a lighting compensation method is used to reduce the illumination sensitivity of the face detection process. Two types of filters, a skin color filter and a neural network filter, are employed to improve the detection efficiency. We introduce a modified convolutional neural network in which a Gabor filter layer is added to generate the feature maps from the input image. A modified fuzzy min-max neural network model for the pattern classifier is also described in this paper. The model employs a new activation function which has the factors of feature distribution and the weight value for each feature in a hyperbox. The weight factor is adjusted through the learning process so that it reflects the degree of relevance of each feature to a pattern class. Through the experimental results using indoor images, the usefulness of the proposed method is discussed.

Important Links:

Go Back