摘要
针对大多数手势识别算法对于形状变化较大的手势鲁棒性不强的现状,提出了一种基于DTW(Dynamic Time Warping)的手势识别算法。论文采用ASL手势数据集作为实验数据,通过图像预处理得到手势的轮廓,再对手势轮廓中心点到轮廓点的距离和轮廓曲率等特征进行提取,最后利用DTW算法寻找规整路径的方法进行识别。实验结果表明,利用DTW算法进行手势识别具有较高的准确率和鲁棒性,识别一幅图像中的手势平均时间小于0.1s,适合于实时手势识别。
To address the shortcoming of most gesture recognition algorithm being with low robustness for evidently deformed gesture,a dynamic time warping(DTW)based gesture recognition algorithm is proposed in this paper. The ASL data sets are employed as testing data. Firstly,the gesture outline is obtained by a pre-processing algorithm. Then gesture feature is extracted,including distance between center point and outline point,and outline point curvity,etc. Finally,gesture is recognized by searching warping path using DTW algorithm. The experimental results show that the gesture recognition algorithm by DTW is with both high recognition rate and high robustness. The average computational time cost for recognition a gesture image is less than 0.1 second,thus being suitable for real-time gesture recognition.
作者
佟喜峰
樊鑫
TONG Xifeng;FAN Xin(School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318)
出处
《计算机与数字工程》
2022年第8期1782-1786,共5页
Computer & Digital Engineering
基金
黑龙江省自然科学基金项目(编号:F2016002)
东北石油大学研究生教育创新工程项目(编号:JYCX_11_2020)资助。
