Gesture recognition based on BoF and its application in human-machine interaction of service robot

The algorithm is applied to the field of hand gesture recognition by Referencing the Bag of Feature (BoF) algorithm in the field of target recognition and image retrieval. First of all, this paper uses the HSV skin color adaptive method to segment the gesture's information from body and uses SURF algorithm to extract the feature of the image. After feature extraction, it uses BoF algorithm to generate the SURF feature visual histogram and mapping to the visual dictionary to generate a unified BOF vector. Then use SVM to train and classify the vector to identify 26 English gestures. This paper selects six gestures as a service robot control commands, which achieve the goal of gesture information control robot. Experimental results show that the BoF algorithm not only has a high time efficiency and accuracy, and meets the real-time performance. The algorithm can achieve real-time control of the robot to stop, start, forward, backward, turn left, turn right, and the accuracy rate is above 90%.