Face detection based on skin color detection and parallel neural network

The purpose of this paper is to detect small faces from crowd of people, and we present face detection method by skin color detection and parallel neural network. Skin color detection is used to detect facial candidate areas. Facial candidate areas are examined by parallel neural network. The parallel neural network is consisted of two networks. The output of them is combined to robustly detect faces. They are trained based on bootstrap method with skin color detection. Simulation results showed that proposed method achieved 71.7% recall rate for 6 × 6 pixel faces, and detected faces from crowd of people with overlapping.