Face Detection by Neural Networks Based on Invariant Moments

摘要

Neural networks beginnings were in the forties, theirs early renaissance experienced in the eighties and today are widely used in pattern recognition, analyzes of medical tests, handwriting, speech, prediction market prices, criminological military research, psychiatric evaluations, data analysis, finding optimal solutions, robot control, weather forecast and many other areas. The great potential of neural networks lies in the possibility of parallel processing of data. Perhaps it will be theirs period of renaissance. This paper introduced an experimental evolution of the effectiveness of utilizing moment invariants as pattern features in human face technology. In this paper we used skin detection and Hu moments for human face localization, also we used neural networks as classifier for this application. The utilized network is a multilayer perceptron (MLP) with one hidden layer. The backpropagation learning is used for its training. Experimental results of neural networks demonstrated successful detection.