Pedestrian detection method is the highest priority for "active safety" which prevents traffic accidents before happens. In previous studies, edge orientation based feature descriptors are proposed. Recently, high standard detection algorithm, Co-occurrence Histograms of Oriented Gradients (CoHOG) is proposed. However, this method has miss detection in complicated situation and processing cost is high. We propose symmetrical judgment algorithm and an extended version of CoHOG for high speed and high accuracy pedestrian detection. The effectiveness of the proposed method was proved on pedestrian detection performance test.