摘要
针对现有方法在动作识别方面存在识别精度不高的缺点,提出一种改进的AdaBoost算法来提高识别的准确率.原算法的弱分类器在训练完后会得到固定的权值系数,改进后的算法在保留所有全部分类器的基础上,在预测阶段加入能度量待测样本与各个弱分类器适应性的动态权值.实验结果表明,改进算法通过固定权值和动态权值乘积与弱分类器进行线性组合,形成最后的强分类器,在手臂动作识别的精度上比原有算法有所提高.
In view of the disadvantages of the current methods in terms of low recognition accuracy,an improved AdaBoost algorithm is proposed to improve the recognition accuracy.The weak classifiers of the original algorithm will obtain fixed weight coefficients after training,and the improved algorithm retains all classifiers and adds the dynamic weight that can measure the adaptability of the test sample and weak classifier in the prediction phase.The experimental results show that the improved algorithm linearly combines the fixed weight and dynamic weight product with the weak classifier to form the final strong classifier,which improves the accuracy of arm action recognition compared to the original algorithm.
作者
杨叶梅
朱秀娥
YANG Yemei;ZHU Xiu'e(Concord University College Fujian Normal University,Fuzhou,Fujian 350117,China)
出处
《福建师大福清分校学报》
2020年第2期26-31,共6页
Journal of Fuqing Branch of Fujian Normal University
基金
福建省中青年教师教育科研项目(JAT170866,JA15634).
