针对证据理论能将多源数据有机合成为具有更高可信度结果的特点,提出基于证据理论融合的手势识别方法.方法先采用Leap Motion采集手势视频序列,提取手势运动轨迹的方向角作为特征;采用隐马尔科夫模型和支持向量机分别对手势进行训练,进而在识别中通过证据理论将两种方法所计算的手势基本概率分配进行决策融合,实现最终的手势识别;将该方法应用于医疗可视化系统中,实现了自然直观的手势交互.实验结果表明,该方法结合了.隐马尔科夫模型和支持向量机的优点,可有效提高手势识别率和交互准确性.%Based on the evidence theory that the multi-source data can be organically synthesis to the higher and more reliable results,the gesture recognition method based on evidence theory fusion is proposed.Firstly,gesture video sequences are captured by Leap Motion and the direction angle of gesture trajectory is extracted as the feature.Secondly,Hidden Markov Model and Support Vector Machine are used for gesture training,and then the basic probability assignment of gesture of the two methods are fused based on decision by evidence theory in recognition processing to realize the final gesture recognition.Finally,this method is applied into medical visualization system to realize the natural and intuitive gesture interaction.The experimental results show that this method combines the advantages of Hidden Markov Model and Support Vector Machine that can effectively improve the gesture recognition rate and the interactive accuracy.
展开▼