多层分类器模型的相似人体活动识别  

Similar Human Activity Recognition Based on a Multi-layer Classification Model

作　　者：李辉[1] 李瑞祥[1] 张耀威 乐燕芬[1] 施伟斌[1] LI Hui;LI Rui-xiang;ZHANG Yao-wei;LE Yan-fen;SHI Wei-bin(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《小型微型计算机系统》2021年第4期861-867,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(51705324)资助。

摘　　要：针对基于单传感器活动识别中相似活动易混淆的问题,本文提出了一种基于广义判别分析的多层分类器融合的相似人体活动识别算法.首先提取基于单加速度计的多类活动数据的时域特征、频域特征以及时频特征,对不同特征进行特征分析与重要性评估以确定有效的特征维度.使用随机森林(RF,Random forest)算法对活动特征进行第1层分类,然后根据分类混淆矩阵分析相似活动,由广义判别分析算法提取相似人体活动的映射特征,使用支持向量机(SVM,Support vector machine)算法对相似活动进行第2层分类,最后将相似活动的双层分类器识别概率加权融合得到最终识别结果.为了验证该识别算法,在公开的数据集SCUT-NAA上执行,识别算法对相似活动识别的正确率达到97.2%,提高了基于该数据集研究的正确率.Aiming at the problem that similar activities recognition based on single sensor are prone to be confused,this paper proposes a recognition algorithm for similar human activities based on a multi-layer classifiers of generalized discriminant analysis. First,the model extracts time-domain features,frequency-domain features,and time-frequency features of different types of activity data based on a single accelerometer,and performs feature analysis and importance evaluation on different features to determine effective feature dimensions. RF( Random forest) algorithm is used to classify the activity features at the first layer. The similar activities are then analyzed according to the confusion matrix,and the transformation characteristics of similar activities are extracted by the generalized discriminant analysis. SVM( Support vector machine) algorithm is used for the second layer of classification of similar activities. Finally,the combined weights of the recognition probabilities of the two-layer classifier of similar activities are used to obtain the final recognition result. We use the public data set SCUT-NAA to test the recognition algorithm. The accuracy rate of this algorithm for similar activities reached 97.2%,showing an improvement of the accuracy rate over previous work based on the same data set.

关 键 词：多层分类器 广义判别分析 活动识别 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]