Welcome to HASCA2018 Web site!

HASCA2018 is a sixth International Workshop on Human Activity Sensing Corpus and Applications: Towards Open-Ended Context Awareness. The workshop will be held in conjunction with UbiComp2018.

Abstract

The recognition of complex and subtle human behaviors from wearable sensors will enable next-generation human-oriented computing in scenarios of high societal value (e.g., dementia care). This will require large-scale human activity corpuses and much improved methods to recognize activities and the context in which they occur. This workshop deals with the challenges of designing reproducible experimental setups, running large-scale dataset collection campaigns, designing activity and context recognition methods that are robust and adaptive, and evaluating systems in the real world. We wish to reflect on future methods, such as lifelong learning approaches that allow open-ended activity recognition.

Unique this year, HASCA will welcome papers from participants to the Sussex-Huawei Locomotion and Transportation Recognition Competition in a special session.

"Sussex-Huawei Locomotion and Transportation Recognition Competition": http://www.shl-dataset.org/activity-recognition-challenge/

The objective of this workshop is to share the experiences among current researchers around the challenges of real-world activity recognition, the role of datasets and tools, and breakthrough approaches towards open-ended contextual intelligence. We expect the following domains to be relevant contributions to this workshop (but not limited to):

Data collection / Corpus construction

Experiences or reports from data collection and/or corpus construction projects, such as papers describing the formats, styles or methodologies for data collection. Cloud- sourcing data collection or participatory sensing also could be included in this topic.

Effectiveness of Data / Data Centric Research

There is a field of research based on the collected corpus, which is called “Data Centric Research”. Also, we solicit of the experience of using large-scale human activity sensing corpus. Using large-scape corpus with machine learning, there will be a large space for improving the performance of recognition results.

Tools and Algorithms for Activity Recognition

If we have appropriate and suitable tools for management of sensor data, activity recognition researchers could be more focused on their research theme. However, development of tools or algorithms for sharing among the research community is not much appreciated. In this workshop, we solicit development reports of tools and algorithms for forwarding the community.

Real World Application and Experiences

Activity recognition "in the Lab" usually works well. However, it is not true in the real world. In this workshop, we also solicit the experiences from real world applications. There is a huge gap/valley between "Lab Envi- ronment" and "Real World Environment". Large scale human activity sensing corpus will help to overcome this gap/valley.

Sensing Devices and Systems

Data collection is not only performed by the "off the shelf" sensors. There is a requirement to develop some special devices to obtain some sort of information. There is also a research area about the development or evaluate the system or technologies for data collection.

Mobile experience sampling, experience sampling strategies:

Advances in experience sampling ap- proaches, for instance intelligently querying the user or using novel devices (e.g. smartwatches) are likely to play an important role to provide user-contributed annotations of their own activities.

Unsupervised pattern discovery

Discovering mean- ingful repeating patterns in sensor data can be fundamental in informing other elements of a system generating an activity corpus, such as inquiring user or triggering annotation crowd sourcing.

Dataset acquisition and annotation through crowd-sourcing, web-mining

A wide abundance of sensor data is potentially in reach with users instrumented with their mobile phones and other wearables. Capitalizing on crowd-sourcing to create larger datasets in a cost effective manner may be critical to open-ended activity recognition. Online datasets could also be used to bootstrap recognition models.

Transfer learning, semi-supervised learning, lifelong learning

The ability to translate recognition mod- els across modalities or to use minimal supervision would allow to reuse datasets across domains and reduce the costs of acquiring annotations.

---