An adaptive model to support biofeedback in AmI environments: a case study in breathing training for autism

Biofeedback systems have shown promising clinical results in regulating the autonomic nervous system (ANS) of individuals. However, they typically offer a “one-size-fits-all” solution in which the personalization of the stimuli to the needs and capabilities of its users has been largely neglected. Personalization is paramount in vulnerable populations like children with autism given their sensory diversity. Ambient intelligence (AmI) environments enable creating effective adaptive mechanisms in biofeedback to adjust the stimuli to each user’s performance. Yet, biofeedback models with adaptive mechanisms are scarce in the AmI literature. In this paper, we propose an adaptive model to support biofeedback that takes the user’s physiological data, user’s adherence to therapy, and environmental data to personalize its parameters and stimuli. Based on the proposed model, we present EtherealBreathing, a biofeedback system designed to help children with autism practice box breathing. We used the data from 20 children with autism using EtheralBreating without adaptation mechanisms to feed an adaptive model that automatically adapts the visual and audible stimuli of EtherealBreathing according to changes in each user’s physiological data. We present two scenarios showing how EtherealBreathing is capable of personalizing the stimuli, difficulty level, or supporting the therapist decisions. Results are promising in terms of performance and personalization of each user model, showing the importance of personalization for AmI technology. Finally, we discuss challenges and opportunities in using adaptive models to support biofeedback in AmI environments.