Intelligent Assistive System Using Real-Time Action Recognition for Stroke Survivors

Cog Watch is an EU project developing technologies for cognitive rehabilitation of stroke patients. The Cog Watch prototype is an automatic system to re-train patients with Apraxia or Action Disorganization Syndrome (AADS) to complete activities of daily living (ADLs). This paper describes the approach to automatic planning based on a Markov Decision Process, and real-time action recognition (AR) based on instrumented objects using Hidden Markov Models. The experimental results demonstrate the ability of a psychologically plausible planning system integrated in a Task Model (TM) to improve task performance via user simulation, and the viability of the approach to AR.