An Exploration on Data Mining for Face Detection based on Real time Face Tracking

Abstract

Abstract
Data mining has been extensively used to gather meaningful information and to improve the significant relationship for the variables warehoused in large data stores. Machine learning provides the technical basis of data mining. Automatic face recognition research which try to give the computer ability to recognize face to distinguish characters. As a key technology of biometrics face recognition technologies, in public security, information security, financial, and other fields has potential application prospect. This paper studies a system of face detection and recognition system based on real-time video. In the face detection part, proposed the Adaboost-ASM face detection algorithm, this algorithm is Adaboost face detection algorithm combined with the Active Shape Model, mainly to solve the problem of the Adaboost face detection algorithm is easy to false drop face in complex region and a similar face region, implement face real-time detection and eliminate non-face region. In the part of face recognition, this paper using 2D-Gabor wavelet feature extraction, and analyzes the advantage and disadvantages of PCA method and Fisher linear discriminant method, and decide using Fisherface method combined with the PCA and Fisher linear discriminant analysis method to reduce the face feature, and then do the projection of the best classification for facial feature. In the end, calculate the projection value of face and projection value of the training sample by cosine distance formula, identifying face. The aim of this research is to implement a face recognition system that will be incorporated to build a system with enhanced properties of the detection algorithm and comparison algorithm for real-time tracking with the basis of logical analysis of data.

Keywords: Data mining, Adaboost, ASM face detection, Gabor feature extraction, PCA, Fisherface, face recognition, Camshift