TSR (Traffic Sign Recognition), a part of ADAS (Advanced Drive Assistance System), helps driver (or car) to recognize traffic signs ahead with using front camera. According to EURO NCAP rating policy, car should be able to warn the driver when the car’s speed is above the set speed threshold. It is thought that various types of traffic sign should be recognized to get more detailed information of road. In this paper, 6 types of traffic sign images are trained by LeNet-5 convolutional neural network architecture. In the detection phase, light-weight color-based segmentation algorithm and Hough transform algorithm are applied to extract candidate regions of traffic signs. The recognition system nearly achieves real-time performance. On-line recognition test is performed on the KAIST campus road, and the result shows all 16 traffic signs are recognized successfully through the driving. The recognition system is implanted into autonomous vehicle ‘Eurecar’. Different types of traffic signs are trained consistently and development of clustering algorithm is considered as a future work for robust recognition system.