改进的mRmR特征选择方法在人体行为识别中的应用

在基于惯性传感器人体行为识别的研究中，特征选择的作用是挑选相关特征，以提高分类性能，去除冗余特征以降低计算复杂度。针对传统的过滤式特征选择方法仅使用一种相关度量系数筛选特征效果不佳的问题，提出一种改进的基于最大相关与最小冗余(mRmR)准则的特征选择方法。该方法在基于mRmR准则下，采用多种相关度量系数融合的方式，在考虑分类类别的条件下，分析待挑选特征与已选特征间的相关性对特征筛选可能产生的积极影响，以去除部分冗余、不相关特征，进而得到初选特征子集；然后利用二进制数对筛选后的特征编码，通过遗传算法搜索最优或次优特征子集。分别使用SVM和KNN分类器对7种日常行为进行分类。实验结果表明，与其他几种方法相比，该方法对实验分类的7种行为有最高的总体平均识别精度，通过SVM和KNN分类的各行为总体平均识别精度分别达到了97.02%和95.73%，与传统的mRmR方法相比，分别提高了13.72%和9.92%。

Application of improved mRmR feature selection in human activity recognition

In the research of human activity recognition based on inertial sensors, feature selection aims to gain relevant features in order to improve classification performance and remove redundant features for computational cost. Due to the fact that traditional filter feature selection methods select features by using only one correlation measure, an improved feature selection method based on mRmR is proposed in this paper. In the framework of mRmR, this method firstly eliminates some redundant features and irrelevant features to obtain the preliminary feature subsets by combining multiple correlation measures and considering the possible positive influence on feature selection when analyzing the correlation between the features to be selected and the selected features under consideration of the classification category. Then, this method uses binary code to encode the gained features to search the optimal or suboptimal feature subset by using genetic algorithm. This paper uses SVM and KNN to classify 7 daily life activities, respectively. The experiment results show that compared with other methods, the proposed method has the highest total average recognition accuracy of 7 behaviors. The total average recognition accuracy of 7 behaviors classified by SVM and KNN reaches 97.02% and 95.73% and compared with the traditional mRmR method, it is also increased by 13.72% and 9.92%, respectively.