Talk Abstract:
Human activity recognition (HAR) is an important concern of the wearable computing community and plays a central role in the field’s vision of context-aware applications and interaction. HAR enables implicit activity-driven interactions and allows for automatically providing just in time information or assistance. In general, sensors, which are worn on the body are utilised to capture aspects of movement or a user’s behaviour.
Ideally, by applying machine learning techniques, this sensor data can be automatically analysed offering a real-time classification of the activities that users are engaged in. In the literature we can find a wide variety of creatively applied classification approaches. By contrast, comparatively little systematic research has addressed the problem of feature design, with almost all previous work using heuristically selected approaches.
This presentation is an introduction to the project 'Deep Learning for Activity Recognition', whose main objective is to perform a characterisation of the benefits of representation learning with deep learning techniques in wearable HAR. Deep neural networks have the potential to discover suitable feature representations that do not rely on application-specific expert knowledge, and may significantly reduce feature engineering effort while increasing HAR performance.
Speaker Biography:
Francisco Javier Ordóñez is a Research Fellow and member of the Sensor Technology Research Centre at the University of Sussex. He is currently involved in the project 'Deep Learning for Activity Recognition' supported by a Google Faculty Research Award grant.
He is interested in the development and application of machine learning techniques that can model and recognise human behaviour from sensor data. Specifically, he has been working with pervasive environments, wireless sensor networks and mobile phones, studying and developing algorithms to perform activity and pattern recognition. Ordóñez received his PhD in computer science from Carlos III University of Madrid.
By: Luke Scott
Last updated: Thursday, 11 June 2015
