Collegial activity learning between heterogeneous sensors

Feuz, Kyle D.; Cook, Diane J.

doi:10.1007/s10115-017-1043-3

Cited by 33 publications

(11 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some studies utilise transfer learning strategies such as teacher/learner to automate an adaptation procedure to address the unsupervised approach. In [21], for example, they developed a system for HAR consisting of multiple views (e.g., one view from environmental sensors and another view from mobile sensors) collaborating to recognise unseen activities on an instance transfer level, reducing the dependency of labelled data.…”

Section: Discussionmentioning

confidence: 99%

Literature Review on Transfer Learning for Human Activity Recognition Using Mobile and Wearable Devices with Environmental Technology

Hernández

Lundström²,

Favela

et al. 2020

SN COMPUT. SCI.

View full text Add to dashboard Cite

Activity recognition systems utilise data from sensors in mobile, environmental and wearable devices, ubiquitously available to individuals. It is a growing research area within intelligent systems that aims to model and identify human physical, cognitive and social actions, patterns and skills. They typically rely on supervised machine-learning approaches, in which the cost of gathering and labelling data is high due to the diverse, interleaved and dynamic nature of human behaviour. Transfer learning is an approach in which previously learned knowledge is utilised to model a new but related setting. For instance, it can reuse existing knowledge to recognise activities performed by different types of users, using different sensor technologies and in different environmental conditions. As the adoption of Internet of Thing devices increases, mobile and wearable sensing is becoming pervasive, and more challenging behaviour recognition activities are being tackled. Yet, the availability of more data does not necessarily translate to better recognition models, if these data are not properly labelled. Thus, the importance of taking advantage of transfer learning to advance the field of activity recognition. This literature review summarises the transfer learning techniques and explores the benefits of combining mobile and wearable devices with environmental sensors in support of transfer learning. We also discuss the maturity of transfer learning by analysing the validation method used in the papers reviewed. Overall, 170 selected articles published between 2014 and 2019 were reviewed following the Okali and Schabram methodology. Findings show an increase of 41% of publications when comparing the output of 2019 against the average number of papers published in the previous 5 years (2014-2018). Inertial sensors such as accelerometers and gyroscopes, are the most frequently used. Feature and instance representation are mature techniques for transfer knowledge. Unsupervised learning across users is a typical application, and shallow techniques and active learning are areas of opportunity in transfer learning methodologies.

show abstract

Section: Discussionmentioning

confidence: 99%

Literature Review on Transfer Learning for Human Activity Recognition Using Mobile and Wearable Devices with Environmental Technology

Hernández

Lundström²,

Favela

et al. 2020

SN COMPUT. SCI.

View full text Add to dashboard Cite

show abstract

“…Thus it is not feasible in most activity cases. Feuz et al [38] proposed a heterogeneous transfer learning method for HAR, but it only learns a global domain shift.…”

Section: Transfer Learning Based Activity Recognitionmentioning

confidence: 99%

Stratified Transfer Learning for Cross-domain Activity Recognition

Wang

Chen

et al. 2018

2018 IEEE International Conference on Pervasive Computing and Communications (PerCom)

183

106

View full text Add to dashboard Cite

In activity recognition, it is often expensive and time-consuming to acquire sufficient activity labels. To solve this problem, transfer learning leverages the labeled samples from the source domain to annotate the target domain which has few or none labels. Existing approaches typically consider learning a global domain shift while ignoring the intra-affinity between classes, which will hinder the performance of the algorithms. In this paper, we propose a novel and general cross-domain learning framework that can exploit the intra-affinity of classes to perform intra-class knowledge transfer. The proposed framework, referred to as Stratified Transfer Learning (STL), can dramatically improve the classification accuracy for cross-domain activity recognition. Specifically, STL first obtains pseudo labels for the target domain via majority voting technique. Then, it performs intra-class knowledge transfer iteratively to transform both domains into the same subspaces. Finally, the labels of target domain are obtained via the second annotation. To evaluate the performance of STL, we conduct comprehensive experiments on three large public activity recognition datasets (i.e. OPPORTU-NITY, PAMAP2, and UCI DSADS), which demonstrates that STL significantly outperforms other state-of-the-art methods w.r.t. classification accuracy (improvement of 7.68%). Furthermore, we extensively investigate the performance of STL across different degrees of similarities and activity levels between domains. And we also discuss the potential of STL in other pervasive computing applications to provide empirical experience for future research.

show abstract

“…Then two classifiers can be trained separately on each view, and then results from each classifier are used to enlarge the training set of the other. The co-training algorithm and its variations have been recently applied in the multi-view human activity recognition in smart home environments [6]. That is, an activity can be viewed from different platforms of sensor streams, such as acceleration data, motion sensor, or video.…”

Section: Co-trainingmentioning

confidence: 99%

SLearn: Shared learning human activity labels across multiple datasets

2018

2018 IEEE International Conference on Pervasive Computing and Communications (PerCom)

View full text Add to dashboard Cite

The research of sensor-based human activity recognition has been attracting increasing attention over years as it is playing an important role in various human-beneficiary applications such as ambient assistive living, health monitoring, and behaviour changing. Nowadays, the advancement of sensing and communication technologies has led to the possibility of collecting a large amount of sensor data, however, to build a reliable computational model and accurately recognise human activities we still need the annotations on sensor data. Acquiring high-quality, detailed, continuous annotations is a challenging task. In this paper, we explore the solution space on sharing annotated activities across different datasets in order to enhance the recognition accuracies. We have designed and developed two approaches: sharing training data and sharing classifiers towards addressing this challenge. We have validated the approach on three datasets and demonstrated their effectiveness in recognising activities only with annotations from as little as 0.1% of each dataset.

show abstract

Collegial activity learning between heterogeneous sensors

Cited by 33 publications

References 65 publications

Literature Review on Transfer Learning for Human Activity Recognition Using Mobile and Wearable Devices with Environmental Technology

Literature Review on Transfer Learning for Human Activity Recognition Using Mobile and Wearable Devices with Environmental Technology

Stratified Transfer Learning for Cross-domain Activity Recognition

SLearn: Shared learning human activity labels across multiple datasets

Contact Info

Product

Resources

About