华中科技大学学报（自然科学版）Issue(z1)：416-419,4.DOI:10.13245/j.hust.15S1099
基于B参数的改进HMM动态手势识别算法
An improved HMM dynamic hand gesture recognition algorithm based on B parameter
摘要
Abstract
For the current hidden Markov model(HMM)in dynamic gesture recognition has a weak a-bility,the training process was complicated and easy to fall into local optimal solution deficiency while training the B parameter,an improved HMM dynamic hand gesture recognition algorithm based on B parameter was proposed.In the calculation of HMM observation state transition probability,this al-gorithm leaded into the improved support vector machine (SVM)algorithm.The strong classification ability of SVM was used to convert the output into probability through Sigmoid function,and the probability was regarded as the each hidden state of the HMM.The improved algorithm optimized the B parameter of the HMM,which improved the performance of the HMM algorithm and enhanced the recognition rate of the system.The experiment proves that this method can accurately identify the dy-namic gesture track,the average recognition rate increases from 86.1% to 97.4%,and has a strong robustness in the illumination and complex environment.关键词
动态手势识别/人机交互/隐马尔可夫模型/支持向量机(SVM)算法/B 参数Key words
dynamic gesture recognition/human-computer interaction/hidden Markov model (HMM)/support vector machine (SVM)algorithm/B parameter分类
信息技术与安全科学引用本文复制引用
张毅,姚圆圆,罗元..基于B参数的改进HMM动态手势识别算法[J].华中科技大学学报（自然科学版）,2015,(z1):416-419,4.基金项目
重庆市教委科学技术研究项目（KJ130512）． （）
