Patients that recover from ligaments, muscles or bone diseases often lack guidance and motivation when performing exercises at home. This leads to longer therapeutic treatments and lower recovery rates. For people that live outside the city commuting time to therapists on average make up double the time of effective training. At the same time therapeutic supervision is essential for the recovery process. Most home exercise aids come equipped with an overhead of technology being accessible or impractical. Our design challenge was to develop a screenless interaction that allows the user to focus on the rehabilitation process at home, while recieving qualitative feedback.

We utilized user-centered research methods and feedback from users and physical therapists to develop a portable device that can be easily attached to different body parts and used with standard physical therapy equipment. This device fills the gap between expensive, hardware-heavy devices and purely software-focused innovations.

Our solution focuses on rehabilitation and prevention. AVA is a compact, portable device that offers screenless interaction to guide the patient without distraction. It is equipped with simple sensors for motion monitoring and instant vibrotactile feedback. At the core lies the embedded machine learning (ML) model that allows for instant classification and feedback, directly interfaced with the therapist. This allows for constant customization, and adaptation of the training sets. Almost no interaction with an app or additional screen is required, as the therapist and patient can train the device through the performed movement. A flexible mounting solution that is compatible with regular physiotherapeutic equipment makes AVA a platform for rehabilitation and prevention.

To allow for remote use we developed an embedded ML model and a flexible mounting solution, making AVA an adaptable and cost-effective solution for physiotherapeutic recovery and prevention.

AVA is a project based on co-creation and exchange between a variety of specialists. Together with the wearable electronics facilitator CPI Electronics we engineered a double sided PCB that is able to run on board machine learning classifications. In collaboration with ML specialist Aeneas Stankowski we designed the guidelines for an algorithmic platform, capable of learning new training models through movements. While previously trained models are recognized and classified leading to visual and haptic feedback. We met with medicine specialists from Charité Berlin and physiotherapists as well as patients. The application experiment was funded by the European commission as part of the Horizon 2020 initiative.