Robust Real-Time Face Detection Using Hybrid Neural Networks
In this paper, a multi-stage face detection method using hybrid neural networks is presented. The method consists of three stages: preprocessing, feature extraction and pattern classification. We introduce an adaptive filtering technique which is based on a skin-color analysis using fuzzy min-max(FMM) neural networks. A modified convolutional neural network(CNN) is used to extract translation invariant feature maps for face detection. We present an extended version of fuzzy min-max (FMM) neural network which can be used not only for feature analysis but also for pattern classification. Two kinds of relevance factors between features and pattern classes are defined to analyze the saliency of features. These measures can be utilized to select more relevant features for the skin-color filtering process as well as the face detection process.
KeywordsFalse Alarm Rate Face Detection Convolutional Neural Network Pattern Classifier Feature Range
Unable to display preview. Download preview PDF.
- 5.Zhu, Q., Cheng, K.T., Wu, C.T., Wu, Y.L.: Adaptive Learning of an Accurate Skin-Color Model. In: Proceeding of the Sixth IEEE International Conf. on Automatic Face and Gesture Recognitin, vol. 1 (2004)Google Scholar
- 7.Kim, H.J., Ryu, T.W., Nguyen, T.T., Lim, J.S., Gupta, S.: A Weighted Fuzzy Min-Max Neural Network for Pattern Classification and Feature Extraction. In: Proceeding of International Conference on Computational Science and Its Application, Part.4, pp. 791–798 (2004)Google Scholar