A Hand Gesture Detection for Multi-Class Cascade Classifier Based on Gradient

摘要

A novel hand gesture detection method in complex background is presented in this paper, it proposed a multi class cascade structure classification based on Gentle AdaBoost (GAB) and Weighted Linear Discriminant Analysis (wLDA). The training and testing experiments are based on the sample database established myself. Histogram of Oriented Gradient (HoG) features of one pair of blocks are extracted with the random size and random locations. Finally, the trained multi class cascade structure classifier for gesture detection is tested and has effectively realized the detection with the proposed method with high detection accuracy in complex background.