摘要
针对基于超声波的手势识别系统在真实场景下的自动化和杂波干扰问题,提出一套包括手势检测、杂波滤除的手势识别算法。手势检测算法基于动态能量阈值,能够检测手势的产生,自动触发手势识别流程。杂波滤除算法使用目标分割,滤除RDM(range Doppler map)图中的杂波,只提取其中有效目标的特征,用于手势的分类。实验结果表明,该系统能够在手势检测和手势识别两个流程间自动切换,对7种不同手势的识别率为97.9%,在存在干扰的真实应用环境中,杂波滤除算法能够提高识别率。
Aiming at problems of automation and clutter interference for ultrasonic-based gesture recognition system in real-world scenarios,a set of gesture recognition algorithm including gesture detection and clutter filtering was proposed.The gesture detection algorithm based on dynamic energy threshold was used to automatically detect whether there was a gesture and to trigger gesture recognition process.Clutter filtering algorithm based on object segmentation was proposed to filter the clutter in RDM(range Doppler map),and thus only the features of effective objects were extracted for classification.Results of experiments show that the system can automatically switch between gesture detection and recognition processes and recognize 7 different gestures with the accuracy of 97.9%.When applied to real application environments with clutter,the clutter filtering algorithm can improve the recognition rate.
作者
周飞飞
李翔宇
ZHOU Fei-fei;LI Xiang-yu(Institute of Microelectronics,Tsinghua University,Beijing 100084,China)
出处
《计算机工程与设计》
北大核心
2020年第3期821-826,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61604014)。
