BraceletSense: A Personalizable Wearable Platform for State Decoding in Non-Verbal Cerebral Palsy with Concepts for Edge AI and Optical Tremor Amplification

This work presents a functional, open-hardware wearable platform and a conceptual framework designed to augment care for non-verbal individuals with severe cerebral palsy (CP). The realized core is a privacy-first, ESP32-based wristband integrating an inertial measurement unit (IMU) and photoplethysmography (PPG) sensor. It streams data to a configurable smartphone interface over a local Wi-Fi network, operating without internet or cloud dependence. The key innovation in the current implementation is a fully personalizable annotation system: caregivers define and label interactive buttons with states specific to their child's unique behavioral language (e.g., "Pre-Agitation," "Seeking Contact," "Sleep-Wake Transition"). This creates a grounded, personalized dataset on the device. Building on this foundation, we propose two forward-looking conceptual pathways: (1) On-Device Personal ML: Utilizing the annotated data to train a lightweight, local machine learning model that learns to predict the caregiver-defined states, transforming the system from a logger to a predictive aid. (2) Passive Optical Amplification: A novel safety-oriented method to capture micro-movements—a gravity-stabilized mirror floating in a glycerin-filled capsule on the bracelet projects a ceiling spot; a camera tracks its amplified displacements for high-fidelity tremor analysis. This work, extending the Livakoland Framework [1] into pragmatic assistive technology, outlines a vision for evolvable, user-owned tools that digitize caregiver intuition to empower care and enable new forms of longitudinal, home-based motor and state analysis.