基于特征联合和偏最小二乘降维的手势识别

针对以往手势识别研究中更关注识别率而弱化实时性的情况,首次将偏最小二乘降维思想引入手势识别领域,提出一种基于特征联合和偏最小二乘降维的手势识别方法。首先进行手势分割,在此基础上提取手势样本的梯度方向直方图和局部二值模式特征,并将二者进行联合。然后采用偏最小二乘法对手势联合特征进行降维,并将降维后的手势训练样本特征输入到支持向量机中进行分类训练。最后用训练好的支持向量机对降维后的手势测试样本进行识别测试。基于Jochen Triesch手势库及自制手势库的实验结果表明,同已有方法相比,本文所提方法在取得较高手势识别率的同时也取得了较好的实时性。

Gesture recognition based on feature combination and partial least squares dimensionality reduction

As it is known that the research had paid more attention to recognition rate rather than real-time performance in the past, in this paper, the idea of partial least squares dimensionality reduction is introduced to the field of gesture recognition for the first time and a novel gesture recognition approach based on partial least squares and support vector machine is proposed. Firstly, the sample features of histograms of oriented gradients and local binary patterns are extracted and combined based on gesture segmen- tation. Secondly, the partial least squares method is adopted to reduce the dimension of the combined features and the combined features after dimensionality reduction is utilized to train the support vector machine. Finally, the gesture testing samples are tested with the trained support vector machine. Experimental results based on the gestures in Jochen Triesch and self-made gesture database show, compared with the existing methods, the proposed approach can achieve better performance on both recognition rate and real-time.