Bio-signal SNN Analytics
Spike-based intelligence for ECG and neuromorphic sensors
Event-driven ECG analysis pipeline implemented on neuromorphic hardware.Abstract
Spiking neural networks (SNNs) provide a biologically plausible alternative to conventional deep learning by transmitting information through binary spikes. We harness this property to build ultra-low-power biosignal analytics pipelines that can run on neuromorphic chips and always-on wearable devices.
Research Highlights
- Neuron-level realism: LIF and adaptive threshold neurons tuned with cardiology-inspired priors
- Spike encoding: Temporal coding schemes that convert ECG and PPG signals into spike trains without losing diagnostic cues
- Hybrid learning: Co-training with ANN teachers followed by spike-based fine-tuning on neuromorphic hardware
- Interpretability: Mapping spike bursts back to physiological events for clinician review
Outcomes
- Prototype ECG arrhythmia monitor consuming <1 mW on neuromorphic fabric
- Tooling that auto-converts PyTorch models into event-driven graphs compatible with Intel Loihi and custom FPGA accelerators
- Joint publications connecting computational neuroscience theory with deployable SNN analytics
Choongseop Lee
Researcher, LG PRI
His research interests include machine learning and computational neuroscience.
Yuntae Park
Researcher, Infinitree
His research interests include machine learning algorithms, deep learning, and reinforcement learning.
Geunbo Yang
Researcher, KG Steel
His research interests include machine learning algorithms and spiking neural networks.
Jiwoon Lee
MS Student
His research interests include computational neuroscience, signal processing and brain-computer interfaces.
Cheolsoo Park
Professor
His research interests include machine learning, adaptive signal processing, computational neuroscience, and wearable technology.