Description
This research focuses on the detection, identification and tracking of people in indoor environments, their activities and vital signs using different types of sensors, either unobtrusive in the environment (radar, long-range capacitive, infrared, acoustic, ultrasound, ...) or in wearable devices. The application domains include assisted living, medicine, home automation, gaming, safety and security in domestic, industrial or public environments. The objectives are the development or improvement of sensors, in particular the reduction of sensitivity to noise and power consumption, and the accuracy of inference from single sensors or multi-sensor data fusion. Techniques explored include efficient and robust sensor analog front-ends, on-sensor (leaf) and edge signal processing (from DSP techniques to efficient neural network and symbolic paradigms), targeted from microcontrollers to systems-on-chip with FPGA accelerators, efficient secure wireless communications, electromagnetic field analysis and optimization. At the higher end, behavioral inference can benefit greatly from (accelerated) cloud processing, using self-correlation and cross-user correlation of large amounts of data.