Classification of Human Daily Activities Using Ensemble Methods Based on Smartphone Inertial Sensors.

摘要: Increasing interest in analyzing human gait using various wearable sensors, which is known as Human Activity Recognition (HAR), can be found recent research. Sensors such accelerometers and gyroscopes are widely used HAR. Recently, high has been shown the use of sensors numerous applications rehabilitation, computer games, animation, filmmaking, biomechanics. In this paper, classification daily activities Ensemble Methods based on data acquired from smartphone inertial involving about 30 subjects with six different discussed. The walking, walking upstairs, downstairs, sitting, standing lying. It involved three stages activity recognition; namely, signal processing (filtering segmentation), feature extraction classification. Five types ensemble classifiers utilized Bagging, Adaboost, Rotation forest, Ensembles nested dichotomies (END) Random subspace. These employed Support vector machine (SVM) forest (RF) base learners classifiers. evaluated holdout 10-fold cross-validation evaluation methods. performance each was measured terms precision, recall, F-measure, receiver operating characteristic (ROC) curve. addition, also comparison overall accuracy rate between learners. observed that overall, SVM produced better 99.22% compared to RF 97.91% a random subspace classifier.