Robust Gesture Recognition Based on Embedded System
Abstract
To improve portability of current gesture recognition technology, this paper proposes a method of the design and implementation of gesture recognition based on embedded system by combining embedded platform with Open CV. Firstly, image denoising using the weighted mean filtering method, and the morphological open operation is used to obtain the complete contour, and the edge extraction is carried out by the improved Canny operator; Secondly, image segmentation using two-dimensional maximum interclass variance. Then analysis and processing of specifies the feature information of gestures images using Hu distance. Finally, the template matching method is used to recognize the gesture image. The experimental results show that the gesture recognition based on embedded system has strong portability, higher recognition rate and easy implementation.
Keywords
Download Options
Introduction
Gesture recognition is one of the most intuitive forms of human-computer interface. It has gradually become a hot spot in many scientific fields. At present, the research on gesture recognition technology is mainly focused on the two aspects of vision and sensor hardware [1,2,3]. Gesture recognition based on sensor acceleration hardware requires the help of some hardware equipment to complete the operation; such equipment is expensive, not easy to wear. Furthermore, that needs to be used in a specific situation, the limitations are very large [4,5].
For gesture recognition based on visual information, it is only necessary to complete the communication between man and machine through human gesture action without intermediate media. It makes gesture recognition convenient and effective because of no external device. And then it brings users a new interactive experience and freedom. Dynamic gestures are one of the most intuitive and effective approach for human-computer interaction [6,7,8,9]. Therefore, we present gesture recognition based on embedded system in order to better recognize and obtain gesture information. The gestures images are acquired by camera are preprocessed, and then the gesture features are extracted and recognized.
Conclusion
At present, many gesture recognition techniques have been applied to PC machines, resulting in great limitations, poor portability and practicability. Therefore, this paper realizes the embedded gesture recognition system by using the combination of embedded system and image processing technology based on the existing gesture recognition technology. Video stream is collected by USB camera, then different gesture commands are obtained to achieve human-computer interaction through gesture processing in the video stream. Finally, the whole gesture recognition system is verified to ensure that the whole system is real-time and effective.