Combining Neural Networks for Skin Detection

Two types of combining strategies were evaluated namely skin features and classifiers. Several rules applied where the outputs classifiers are combined using binary operators such as AND OR operators, Voting, Sum Weights a new neural network. Three chrominance components from YCbCr colour space that gave highest correct detection on their single feature MLP selected parameters. A major issue in designing network is to determine optimal number hidden units given set training patterns. Therefore, coarse fine search method find neurons layer proposed. The strategy Cb/Cr Cr improved by 3.01% compared best Cb-Cr. three rule further 4.38% MLP.