Ubidenz - Ubiquitous Digital Empathic Therapy Assistance


Projektstart: 01.09.2021
Duration: 36 month
Funded by: BMBF (Bundesministerium für Bildung und Forschung)
Local head of Project: Prof. Dr. Elisabeth André
Local Scientist: Merlin Albes, M.Sc.
© University of Augsburg

About the Project

As a socio-empathic assistance system for people suffering from depression who have already been discharged from hospital, UBIDENZ provides innovative aftercare management and automated monitoring of the further course of outpatient aftercare.

The University of Augsburg is responsible for the real-time analysis of therapy-relevant behaviors in the UBIDENZ project. The automated and continuous categorization of emotional states based on machine learning and artificial intelligence provides a detailed analysis of emotions over longer periods of time and thus allows important conclusions to be drawn about the psychological state and possible effects of therapeutic measures. Persistent negative mood states, lack of emotional expressivity, and lack of activation, for example, may indicate depressive disorders.

Recognition of the patient's intrapersonal state is a fundamental component of the understanding that the UBIDENZ avatar needs in order to create an empathetic interaction with the patient. Furthermore, however, it is necessary to analyze physiological and social signals within the human-machine interaction. In addition to the intrapersonal analysis of the patient, the UBIDENZ project develops algorithms for the detection and analysis of interpersonal behavioral signals between patient and avatar.

The automated coding software Nonverbal Behaviour Analyzer (NOVA) will be continuously enhanced during the course of the study based on the findings from the practical application. In particular, the user-friendliness and comprehensibility of the functions will be improved. In order to maximize the acceptance and usability of the developed system, the UBIDENZ project integrates interactive and multimodal explanation components, which make the decisions of the automatic recognition models more transparent, comprehensible and trustworthy.