On October 30th, Miguel defended his PhD thesis entitled Human Activity Recognition with Consumer Devices and Real-Life Perspectives” under the supervision of Carlos Granell and Sven Casteleyn. This thesis is also connected with the research activities under the SyMptOMS-related projects, a series of research projects that marry the technical expertise of the GeoSpatial Technologies Research Lab (GEOTEC) with the knowledge in experimental psychology of the Laboratory of Psychology and Technology (LABPSITEC) to develop innovative, mobile-based solutions for the diagnosis, treatment, and relapse prevention of mental health disorders.

Miguel’s work focuses on Human Activity Recognition (HAR) using consumer devices such as smartphones and smartwatches. First, a comprehensive analysis was performed to compare the performance of different machine learning (ML) and deep learning (DL) models based on the amount of training data, data source, and model architecture.

Second, the thesis also presents an mHealth system to automate the Timed Up and Go (TUG) test, which is used to assess mobility. The evaluation of the system demonstrated its accuracy, reliability, and low power consumption, making it a promising solution for self-administration of TUG tests.

Finally, the thesis explores the feasibility of HAR systems based on Wi-Fi channel state information (CSI), using ESP32 microcontrollers. While initial results showed promising performance, further experiments revealed that the instability of CSI data over time poses a significant challenge for real-life applications.

Overall, Miguel’s thesis provides a significant contribution to the HAR field by addressing key challenges and exploring new avenues for the development of real-life HAR systems. You deserve it!