一种基于深度图像的静态手势神经网络识别方法

静态手势识别是以手势驱动的人机交互系统的核心技术。针对静态手势识别问题,提出了一种基于深度图像进行静态手势识别的方法。为了消除静态手势识别过程中的平移、旋转和缩放不变性,提取手势轮廓的Hu不变矩,并以Hu不变矩作为特征构建静态手势深度感知神经网络模型,以此实现对静态手势进行分类识别。在VisualStudio的开发环境下实现了对该方法的验证,取得了良好的效果,并与传统的模板匹配法与基于卷积神经网络的深度学习方法作比较,静态手势识别准确率总体可达95%,识别效率高,能满足实时性要求。

A static gesture recognition method based on depth image and neural network

Static gesture recognition is the key technology of human-computer interaction system driven by gesture.To solve the problem of static gesture recognition,a method of static gesture recognition based on depth image is proposed.In order to eliminate the invariance of translation,rotation and scaling in the process of static gesture recognition,the Hu moment of gesture contour is extracted based on Hu moment theory,and the static gesture depth perception neural network model is constructed with Hu moment as the feature.In this way,static gestures are classified and recognized.The method is validated under the development environment of VisualStudio,and the result is fine.It is compared with the traditional template matching method and the depth learning method based on convolution neural network.The accuracy of static gesture recognition in this paper is high,and the recognition efficiency is high and it can meet the real-time requirements.