基于视觉的静态手势识别系统

摘要

为了丰富手势识别方法的多样性, 提高手势识别的正确率, 提出了一种基于手势轮廓像素变化的手势识别方法.在Matlab环境下, 设计并开发了一个基于视觉的静态手势识别系统.系统主要由两部分组成:手势分割与手势识别.该系统通过摄像头实时采集手势图像, 根据HSV颜色空间上的聚簇特性, 通过结合RGB和HSV双颜色空间的手势分割方法得到二值化图像, 再对图像进行平滑滤波、形态学处理等处理后得到矩形手势区域图像.手势轮廓可以很好地表征手势图像的局部特征, 文中在获取手势轮廓的基础上根据手势区域的像素变化进行手势识别.实验结果表明, 该系统能够对规定的四种常用静态手势进行识别, 系统鲁棒性很好, 平均识别率达到了90％, 总体识别有效率达到了85.9％, 有较好的识别效果.%In order to enrich the diversity of gesture recognition method and improve the accuracy of gesture recognition, we propose a gesture recognition method based on the change of gesture contour pixel.In the Matlab environment, we design and develop a visual static gesture recognition system which consists of two parts:gesture segmentation and gesture recognition.This system collects the gesture images in real time through the camera.According to the clustering characteristics of HSV color space, the binary image is obtained by combining the gesture segmentation method of RGB and HSV dual-color space, and then the image is processed by smoothing and morphological processing.The gesture contour may well characterize the local feature of the gesture image, which is signed in response to the pixel change of the gesture region based on the acquired gesture contour.The experiment shows that the system can identify four commonly used static gestures with strong robustness and great recognition effect.The average recognition rate is up to 90％ and the overall recognition rate is 85.9％.