Research On Gesture Recognition Method Based On Depth Learning

Gesture recognition is an important research topic in human-computer interaction.Because the research on it,especially on visual-based gesture recognition,conforms to the trend of human-computer interaction from machine-friendly to human-friendly development in recent years,and therefore has great Research and application prospects.However,in the actual use,the diversity of human form,and its background in the environment,changes in light and other factors give the computer great challenges to correctly identified gestures from the image information.In view of these problems,this paper has studied the gesture recognition,gesture detection and other issues,the main work is as follows:(1)For the gesture detection problem,we combined with a variety of video detection algorithm proposed a multi-strategy fusion gesture detection method.In order to solve the problem of false detection in gesture detection in complex background,the principle of HSV skin color detection and vibe motion detection is studied.According to the characteristics of various algorithms in the detection,the information of skin color and moving face is merged to improves the robustness of gesture detection in complex background.In particular,The fusion strategy used the adaptive threshold,which improves the detection rate of the algorithm in the case that gestures coincide with skinny background.(2)A gesture recognition method based on binary convolution neural network is proposed based on the general depth learning convolution neural network gesture recognition method for gesture classification recognition problem.This method combines the binarization method of the network with the convolution neural network gesture recognition method,and uses the binarized weight to replace the original high precision weight in the network,which reduces the computational cost and memory occupancy.Experiments show that the algorithm has achieved sufficient accuracy and robustness,and the computational efficiency and applicability in real-time system have beenimproved.(3)The design and implementation of a gesture recognition system,showing the gesture recognition in the human-computer interaction system applications.From the design of the system's demand and function modules to the realization of the gesture recognition function module and the gesture training module in the complex background combining the two methods proposed above,and then the implementation of collaborative authentication module integrating the mature face recognition detection scheme,the article describes in detail the system design and implementation of the various details.Finally,the experiment shows the function and characteristics of the system used to identify the number and unlock.