NEUROMORPHIC COMPRESSED TELEMETRY FOR REAL-TIME, LOW-POWER WIRELESS TRANSMISSION OF HIGH-DENSITY NEURAL RECORDINGS

Brain-computer interfaces (BCIs) can restore or augment function for people with neurological disorders, but scalable wireless systems are limited by the power and safety cost of transmitting high-density recordings that capture single-unit activity. This study aims to reduce telemetry bandwidth and energy while preserving waveform fidelity required for downstream analyses such as spike sorting. We introduce Neuromorphic Compressed Telemetry (NCT), a real-time compression system with two stages: temporal delta modulation that encodes sample-to-sample differences, followed by a biologically inspired spiking neural network (SNN) that selectively represents salient features, including action potentials. NCT supports training mode to prioritize high-fidelity waveform reconstruction, compress mode to maximize real-time efficiency, and monitor mode to track longer-term signal changes and adapt to biological variability such as electrode drift. An event-based transmission protocol packages mode outputs into compact messages for wired or wireless links. Evaluated on real neural recordings, NCT achieves more than 80× data reduction while maintaining signal quality comparable to the noise floor of typical recording systems and preserving information needed for spike sorting. A compact integrated circuit measuring 1.5 by 3 mm² processes 384 channels while consuming less than 1 mW. These results indicate that NCT can enable scalable, battery-limited wireless BCIs that record from multiple brain areas without tissue overheating or bulky external hardware. The multi-mode design allows rapid retraining when conditions change and continuous monitoring during chronic implantation.