With the continuous development of technology and innovation,using driving simulator for driving training gradually becomes increasingly popular.Driving simulator is not only safe and convenient tools for whoever with or without driving experience,but also reaches the sensation like using the real car in training.However,the existing driving simulator has a drawback which does not feed any information back to users especially driving postures.The advantages of driving with correct driving postures can improve the efficiency of car control,reduce the unexpected accident rate and reduce ache or fatigue too.In order to overcome this shortcoming,this dissertation studies the design and implementation of Driver Behavior Training Expert System based on driving posture recognition.The purpose of this study aims to train the driving learner who lacks the correct driving knowledges including correct driving postures and good driving habits.This study considers the driving postures in relation with head movement,hands and feet activities in order to recognize the driver posture.A rule-based expert system is built to evaluate whether those driving postures are correct by analyzing the series of postures in real-time.If incorrect driving posture or bad driving habit has occurred,this proposed method can correct them by giving the proper advice to driving learner via automatic voice prompt.In summary,the main innovations and contributions of this dissertation are listed as the following:（1）Driving Postures and Driving Habits Static Recognition——this dissertation presents how the proposed system can recognize the driving learner behavior including head pose,hands,feet operating,fasten seat belt,footwear and also driving habits in static model.Skin color segmentation based on YCbCr color space model is used to e0078tract the skin-like area and non-skin like area,then Support Vector Machine（SVM）Classification with Histogram of Oriented Gradient（HOG）feature is applied to classify the driving posture.The classification result shows that SVM classification with HOG feature can recognize the driving postures and driving habits with high precise and high accuracy rate,the more data sets,the more accuracy rate we got.（2）Pose and Habit Dynamic Detection——this proposed system enables to detect the novice driver posture and habit including head pose estimation,hands operation on steering wheel and gear shift,pedals operation and reckless driving posture,fasten seat belt and type of shoes based on safety driving in dynamic model.Fuzzy logic and Perspective-n-Point are applied to form the head posed estimation.The rest used the simple skin color segmentation as pre-processing and applied the recognition to detect the pose.Various image processing methods i.e.,image blurring or morphological operation are used to enhance the quality of images and to decrease noise within images.The results submit the effectiveness of the proposed technique in detection the novice driver’s posture and habit.（3）Rule-Based Expert System——this proposed system enables to evaluate and analyze whether those driving postures and habits are correct.The sequence rules of driving postures and habits are collected in knowledge base（KB）and are encoded by using forward chaining inference engine and finite state machine（FSM）is used to design the driving postures and driving assistance system in real-time.As the result shows the robustness of this proposed method,this system can evaluate and give the proper advice for driving learner according to the correct driving posture and good driving habit.