Name
Abstract | ||
|---|---|---|
In the last years, Activity Recognition (AR) has drawn the attention of many researchers in several fields such as user mobility identification and monitoring of patients and of daily activities. In the context of Dead Reckoning (DR) navigation systems, AR has been used so far to get landmarks on the map of the buldings and permit the calibration of the considered routines. The present work aims at providing a contribution to the definition of a more effective recognition of movement in the DR applications. To this aim we describe the implementation of a Movement Segmentation procedure which permits to distinguish between posture change movements, such as lying down and standing up, and cyclic movements such as walking, walking downstairs and upstair. As it is known, these movements which are very similar and prone to critical recognition analysis, can often be misleaded; therefore, they are considered as inputs of a supervised learning technique which allows their classification. Particularly, the acceleration data are acquired from a Motion Node sensor that is worn on front right-hip and four supervised learning classification families, namely the Decision Tree (DT), the Support Vector Machine (SVM), the K-Nearest Neighbor (KNN) and the Ensamble Learner (EL), are tested. The accuracy of the considered classification models is evaluated; particularly, the confusion matrices are presented which shed light on the collection of the movements that are more likely to be mixed up. |
Year | DOI | Venue |
|---|---|---|
2015 | 10.1109/IPIN.2015.7346769 | 2015 International Conference on Indoor Positioning and Indoor Navigation (IPIN) |
Keywords | Field | DocType |
daily movement recognition,dead reckoning applications,activity recognition,user mobility identification,patients monitoring,daily activities,DR navigation systems,landmarks,DR applications,movement segmentation,posture change movements,lying down,standing up,cyclic movements,downstairs walking,upstair walking,supervised learning technique,acceleration data,motion node sensor,supervised learning classification families,decision tree,support vector machine,SVM,k-nearest neighbor,KNN,ensamble learner,classification models,confusion matrices | Computer vision,Decision tree,Activity recognition,Accelerometer,Segmentation,Support vector machine,Feature extraction,Supervised learning,Dead reckoning,Artificial intelligence,Engineering,Machine learning | Conference |
ISSN | Citations | PageRank |
2162-7347 | 2 | 0.40 |
References | Authors | |
9 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Alessio Martinelli | 1 | 15 | 6.01 |
| Simone Morosi | 2 | 105 | 25.84 |
| Enrico Del Re | 3 | 212 | 33.24 |