Energy-Efficient\nReservoir Computing Based on Solution-Processed\nElectrolyte/Ferroelectric Memcapacitive Synapses for Biosignal Classification

The\nclassification of critical physiological signals\nusing neuromorphic\ndevices is essential for early disease detection. Physical reservoir\ncomputing (RC), a lightweight temporal processing neural network,\noffers a promising solution for low-power, resource-constrained hardware.\nAlthough solution-processed memcapacitive reservoirs have the potential\nto improve power efficiency as a result of their ultralow static power\nconsumption, further advancements in synaptic tunability and reservoir\nstates are imperative to enhance the capabilities of RC systems. This\nwork presents solution-processed electrolyte/ferroelectric memcapacitive\nsynapses. Leveraging the synergistic coupling of electrical double-layer\n(EDL) effects and ferroelectric polarization, these synapses exhibit\ntunable long- and short-term plasticity, ultralow power consumption\n(∼27 fJ per spike), and rich reservoir state dynamics, making\nthem well-suited for energy-efficient RC systems. The classifications\nof critical electrocardiogram (ECG) signals, including arrhythmia\nand obstructive sleep apnea (OSA), are performed using the synapse-based\nRC system, demonstrating excellent accuracies of 97.8 and 80.0% for\narrhythmia and OSA classifications, respectively. These findings pave\nthe way for developing lightweight, energy-efficient machine-learning\nplatforms for biosignal classification in wearable devices.