BREAKING THE COST-PERFORMANCE BARRIER: SINGULAR ARCHITECTURE FOR EEG, EMG, AND BIOPOTENTIALS

Translating biosignal technologies from laboratory settings to widespread clinical and commercial use has been hindered by prohibitive costs of research-grade hardware, inflexible single-modality architectures, and integration challenges across fragmented platforms. To address these barriers, we developed Instinct—a modular acquisition system built on a decoupled three-tier architecture comprising electrode-level signal conditioning, regional digitization, and centralized coordination. This design supports simultaneous EEG, ECG, and EMG recording without compromising signal integrity. Neurophysiological benchmarking confirmed research-grade performance: SSVEP and alpha modulation paradigms yielded signals of 10dB and 15dB respectively above baseline, matching established commercial systems.
We validated Instinct for its practical utility in human-computer interaction applications through Project Kinesis. We recorded 8-channel sEMG from forearm using Instinct across a total of 25 participants during a set of up to 27 hand gestures spanning basic (ex. extension, flexion), communicative (ex. fist, victory), directive (ex. pointing), and device-interactive gestures (ex. scrolling, swiping). Our analysis showed higher correlations across gestures that recruited similar set of muscles (ex. victory and pointing index finger), and anti-correlation in anatomically antagonistic gestures (ex. extension and flexion), confirming the validity of the system and protocols. We further recorded approximately 30s seconds of data per user per gesture, and were able to achieve a 7-gesture classification accuracy between 85 and 95% for individual participants, and 60-70% across participants.
Instinct supports scalable configurations: 8 to 128 channels, across modalities (ExG, PPG, GSR) and arbitrary form factors. Hence this platform bridges the gap between fundamental neuroscience investigation, clinical rehabilitation, and next-generation brain-computer interfaces.