ESR4. Continuous behaviour and physiology monitoring using ambient acoustics and thermal images.
Project Title: Continuous behaviour and physiology monitoring using ambient acoustics and thermal images.
Description: Sensors and processing capabilities of modern embedded computers, including mobile, wearable, and implantable devices, can be used to recognise a user's situation and environment continuously, providing context awareness. As sensor modalities and data bandwidth increase, context recognition applications diversify and continuously run statistical pattern recognition algorithms. However, the pattern recognition does not require a constant processing load at all times, i.e., to repeatedly apply sliding windows over equidistant sampled data and sequentially iterate recognition algorithms over the windows. Instead, sampling and processing at relevant moments will be sufficient to obtain acceptable accuracy, where the system could sleep in low power states during the remaining time. To realise a dynamic, on-demand context learning and recognition approach, it is essential to investigate sparse representations and suitable process control strategies that can optimise between accuracy, energy, and latency, while being configurable to support varying services requested by applications.
ESR4: Giovanni Schiboni
Affiliation: Universität Passau, ACTLAb Research Group