| 1. |
- Cleland, Ian, et al.
(författare)
-
Collection of a Diverse, Naturalistic and Annotated Dataset for Wearable Activity Recognition
- 2018
-
Ingår i: 2018 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops). - : IEEE. ; , s. 555-560
-
Konferensbidrag (refereegranskat)abstract
- This paper discusses the opportunities and challenges associated with the collection of a large scale, diverse dataset for Activity Recognition. The dataset was collected by 141 undergraduate students, in a controlled environment. Students collected triaxial accelerometer data from a wearable accelerometer whilst each carrying out 3 of the 18 investigated activities, categorized into 6 scenarios of daily living. This data was subsequently labelled, anonymized and uploaded to a shared repository. This paper presents an analysis of data quality, through outlier detection and assesses the suitability of the dataset for the creation and validation of Activity Recognition models. This is achieved through the application of a range of common data driven machine learning approaches. Finally, the paper describes challenges identified during the data collection process and discusses how these could be addressed. Issues surrounding data quality, in particular, identifying and addressing poor calibration of the data were identified. Results highlight the potential of harnessing these diverse data for Activity Recognition. Based on a comparison of six classification approaches, a Random Forest provided the best classification (F-measure: 0.88). In future data collection cycles, participants will be encouraged to collect a set of “common” activities, to support generation of a larger homogeneous dataset. Future work will seek to refine the methodology further and to evaluate model on new unseen data.
|
|
| 2. |
- Cruciani, Frederico, et al.
(författare)
-
Automatic annotation for human activity recognition in free living using a smartphone
- 2018
-
Ingår i: Sensors. - : MDPI. - 1424-8220. ; 18:7
-
Tidskriftsartikel (refereegranskat)abstract
- Data annotation is a time-consuming process posing major limitations to the development of Human Activity Recognition (HAR) systems. The availability of a large amount of labeled data is required for supervised Machine Learning (ML) approaches, especially in the case of online and personalized approaches requiring user specific datasets to be labeled. The availability of such datasets has the potential to help address common problems of smartphone-based HAR, such as inter-person variability. In this work, we present (i) an automatic labeling method facilitating the collection of labeled datasets in free-living conditions using the smartphone, and (ii) we investigate the robustness of common supervised classification approaches under instances of noisy data. We evaluated the results with a dataset consisting of 38 days of manually labeled data collected in free living. The comparison between the manually and the automatically labeled ground truth demonstrated that it was possible to obtain labels automatically with an 80–85% average precision rate. Results obtained also show how a supervised approach trained using automatically generated labels achieved an 84% f-score (using Neural Networks and Random Forests); however, results also demonstrated how the presence of label noise could lower the f-score up to 64–74% depending on the classification approach (Nearest Centroid and Multi-Class Support Vector Machine).
|
|
| 3. |
- Cruciani, Federico, et al.
(författare)
-
Personalized Online Training for Physical Activity monitoring using weak labels
- 2018
-
Ingår i: 2018 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops). - : IEEE. - 9781538632277 ; , s. 567-572
-
Konferensbidrag (refereegranskat)abstract
- The use of smartphones for activity recognition is becoming common practice. Most approaches use a single pretrained classifier to recognize activities for all users. Research studies, however, have highlighted how a personalized trained classifier could provide better accuracy. Data labeling for ground truth generation, however, is a time-consuming process. The challenge is further exacerbated when opting for a personalized approach that requires user specific datasets to be labeled, making conventional supervised approaches unfeasible. In this work, we present early results on the investigation into a weakly supervised approach for online personalized activity recognition. This paper describes: (i) a heuristic to generate weak labels used for personalized training, (ii) a comparison of accuracy obtained using a weakly supervised classifier against a conventional ground truth trained classifier. Preliminary results show an overall accuracy of 87% of a fully supervised approach against a 74% with the proposed weakly supervised approach.
|
|
| 4. |
- Cruciani, Federico, et al.
(författare)
-
Personalizing Activity Recognition with a Clustering based Semi-Population Approach
- 2020
-
Ingår i: IEEE Access. - : IEEE. - 2169-3536. ; 8, s. 207794-207804
-
Tidskriftsartikel (refereegranskat)abstract
- Smartphone-based approaches for Human Activity Recognition have become prevalent in recent years. Despite the amount of research undertaken in the field, issues such as cross-subject variability are still posing an obstacle to the deployment of solutions in large scale, free-living settings. Personalized methods (i.e. aiming to adapt a generic classifier to a specific target user) attempt to solve this problem. The lack of labeled data for training purposes, however, represents a major barrier. This is especially the case when taking into consideration that personalization generally requires labeled data to be user-specific. This paper presents a novel personalization method combining a semi-population based approach with user adaptation. Personalization is achieved through the following. Firstly, the proposed method identifies a subset of users from the available population as best candidates for initializing the classifier to the target user. Subsequently, a semi-population Neural Network classifier is trained using data from this subset of users. The classifier’s network weights are then updated using a small amount of labeled data from the target user subsequently implementing personalization. This approach was validated on a large publicly available dataset collected in a free-living scenario. The personalized approach using the proposed method has shown to improve the overall F-score to 74.4% compared to 70.9% when using a generic non-personalized approach. Results obtained, with statistical significance being confirmed on a set of 57 users, indicate that model initialization using the semi-population approach can reduce the amount of labeled data required for personalization. As such, the proposed method for model initialization could facilitate the real-world deployment of systems implementing personalization by reducing the amount of data needed for personalization.
|
|
| 5. |
- Nugent, Chris, et al.
(författare)
-
An initiative for the creation of open datasets within pervasive healthcare
- 2016
-
Ingår i: Proceedings of the 10th EAI International Conference onPervasive Computing Technologies for Healthcare. - : ICST, the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering. - 9781631900518 ; , s. 318-321
-
Konferensbidrag (refereegranskat)abstract
- In this paper issues surrounding the collection, annotation, management and sharing of data gathered from pervasive health systems are presented. The overarching motivation for this work has been to provide an approach whereby annotated data sets can be made readily accessible to the research community in an effort to assist the advancement of the state-of-the-art in activity recognition and behavioural analysis using pervasive health systems. Recommendations of how this can be made a reality are presented in addition to the initial steps which have been taken to facilitate such an initiative involving the definition of common formats for data storage and a common set of tools for data processing and visualization.
|
|
| 6. |
- Synnes, Kåre, 1969-, et al.
(författare)
-
H2Al : The Human Health and Activity Laboratory
- 2018
-
Ingår i: <em>Proceedings</em>, 2018, UCAmI 2018. - Basel Switzerland : MDPI.
-
Konferensbidrag (refereegranskat)abstract
- The Human Health and Activity Laboratory (H2Al) is a new research facility at Luleå University of Technology implemented during 2018 as a smart home environment in an educational training apartment for nurses and therapists at the Luleå campus. This paper presents the design and implementation of the lab together with a discussion on potential impact. The aim is to identify and overcome economical, technical and social barriers to achieve an envisioned good and equal health and welfare within and from home environments. The lab is equipped with multiple sensor and actuator systems in the environment, worn by persons and based on digital information. The systems will allow for advanced capture, filtering, analysis and visualization of research data such as A/V, EEG, ECG, EMG, GSR, respiration and location while being able to detect falls, sleep apnea and other critical health and wellbeing issues. The resulting studies will be aimed towards supporting and equipping future home environments and care facilities, spanning from temporary care to primary care at hospitals, with technologies for activity and critical health and wellness issue detection. The work will be conducted at an International level and within a European context, based on a collaboration with other smart labs, such that experiments can be replicated at multiple sites. This paper presents some initial lessons learnt including design, setup and configuration for comparison of sensor placements and configurations as well as analytical methods.
|
|
