A multiple features fusing algorithm is presented to detect pedestrian based on their appearance characteristics. Firstly, the edge, texture and color frequency features are extracted via HOG, Co-occurrence Matrices and HSV color in features-extracting stage, and a richer feature set is formed. Then, some important features are selected via partial least squares for dimension reduction and a quadratic discriminant analysis classifier is formed in classifier-creating stage. Lastly, the pedestrian is monitored by a well trained classifier. Experimental results show that the proposed algorithm outperforms both the HOG and PID algorithm, with a 3% miss rate at 10-4FPPW.