基于傅里叶描述子的手势识别方法  

Gesture Recognition Method Based on Fourier Descriptor

作　　者：邢益良[1] 雷华军[1] 

机构地区：[1]海南软件职业技术学院,海南琼海571400

出　　处：《工业控制计算机》2024年第5期77-79,82,共4页Industrial Control Computer

基　　金：海南省自然科学基金资助项目(621RC607)。

摘　　要：手势识别是计算机视觉人机交互应用领域关键技术,手势轮廓携带有手势重要特征,准确捕获手势轮廓对提高手势识别具有重要意义。针对手掌轮廓特征提取困难和手势识别率低问题,提出了基于傅里叶描述子的手势识别方法,按照候选窗口最大轮廓傅里叶描述子匹配度和置信度分割出手掌区域;跟踪手掌轮廓计算其傅里叶描述子得到手势轮廓特征值;将16个手势轮廓特征值作为BP人工神经网络的输入,利用BP人工神经网络识别手势。实验表明,该方法能有效捕获手势轮廓和识别19种手势,具有识别率高、性能优良和鲁棒性好等优点。This paper proposes a new gesture recognition method based on fourier descriptor to solve the problems of palm contour feature extraction and low gesture recognition rate.Firstly,this paper segments the palm region based on the maximum contour fourier descriptor matching and confidence of the candidate window.Then,this paper tracks the palm contour and calculates its fourier descriptor to obtain the gesture contour feature values.Finally,this paper utilizes the BP artificial neural network to receive the 16 gesture contour feature values and recognize the gesture categories.Experiments have shown that the proposed method can effectively capture gesture contours and recognize 19 categories of gestures,with advantages such as high recognition rate,excellent performance,and good robustness.

关 键 词：手势识别 傅里叶描述子 手势轮廓 人工神经网络 候选窗口 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]