基于三支决策的触摸手势识别算法

Algorithm of touch gesture recognition based on three-way decisions

为了提高人与机器人交互过程中对触摸手势的识别能力,提出一种基于三支决策的触摸手势识别算法。通过对触摸手势数据集Co ST(corpus of social touch)的分析以及结合其他领域的研究,提出2种数据预处理方法"截取"和"去背景",并从6个不同角度提取了特征,包括基本特征、基于直方图的特征、序列特征、梯度特征、接触面积特征和基于每个传感器的特征。以随机森林为分类器采用十折交叉方法进行了验证,发现不同的预处理对于不同触摸手势的识别有不同的效果。为了融合不同预处理的优势,引入三支决策的思想,将m分类问题转化为m个2分类问题,使用统计的方法计算每个二分类的三支决策阈值,按照一定的先后顺序和权重指标对经过不同预处理的分类结果进行决策筛选。仿真实验结果表明,基于三支决策的触摸手势识别算法在一定程度上达到了融合的效果,并提高了触摸手势的识别率。

In order to improve the recognition ability of the touch gesture in the interaction between human and robot,an algorithm of touch gesture recognition based on three-way decisions is proposed.Through the analysis of touch gesture data set Co ST(Corpus of Social Touch) and combining with other fields of study,we put forward two data preprocessing methods," cutout" and " instant mask",and extract features from six different perspectives including basic features,histogrambased features,sequence features,gradient-based features,contact area features and channel-based features.Using random forest as classifier and validating with 10-fold cross method,it is found that different preprocessing methods have different effects on the recognition of different gestures.In order to combine the advantage of different preprocessing,threeway decisions is introduced,and the m-category classification problems are changed into m two-category classification problems.The three-way decisions' thresholds for each two-category classification problem are calculated by statistical method.According to a certain order and indicators,the classification results through different preprocessing are screened.The experimental results show that the algorithm achieves a certain degree of fusion effect and improves the recognition rate of the touch gesture.