Heart-rate estimation is a fundamental feature of modern wearable devices. In this paper we propose a machine learning technique to estimate heart-rate from electrocardiogram (ECG) data collected using wearable devices. The novelty of our approach lies in (1) encoding spatio-temporal properties of ECG signals directly into spike train and using this to excite recurrently connected spiking neurons in a Liquid State Machine computation model; (2) a novel learning algorithm; and (3) an intelligently designed unsupervised readout based on Fuzzy c-Means clustering of spike responses from a subset of neurons (Liquid states), selected using particle swarm optimization. Our approach differs from existing works by learning directly from ECG signals (allowing personalization), without requiring costly data annotations. Additionally, our approach can be easily implemented on state-of-the-art spiking-based neuromorphic systems, offering high accuracy, yet significantly low energy footprint, leading to an extended battery-life of wearable devices. We validated our approach with CARLsim, a GPU accelerated spiking neural network simulator modeling Izhikevich spiking neurons with Spike Timing Dependent Plasticity (STDP) and homeostatic scaling. A range of subjects is considered from in-house clinical trials and public ECG databases. Results show high accuracy and low energy footprint in heart-rate estimation across subjects with and without cardiac irregularities, signifying the strong potential of this approach to be integrated in future wearable devices.

Simultaneous measurement of Electrocardiogram (ECG) and bio-impedance (BioZ) via disposable health patches is desired for patients suffering from chronic cardiovascular and respiratory diseases. However, a sensing IC must consume ultra-low power under a sub-volt supply to comply with miniaturized and disposable batteries. This work presents a 0.6 V analog frontend (AFE) IC consisting of an instrumentation amplifier (IA), a current source (CS) and a SAR ADC. The AFE can measure ECG and BioZ simultaneously with a single IA by employing an orthogonal chopping scheme. To ensure the IA can tolerate up to 300mVpp DC electrode offset and 400mV _pp common-mode (CM) interference, a DC-servo loop (DSL) combined with a common-mode feedforward (CMFF) loop is employed. A buffer-assisted scheme boosts the IA's input impedance by 7x to 140MΩ at 10Hz. To improve the BioZ sensitivity, the CG utilizes dynamic element matching to reduce the 1/f noise of the output current, leading to 35mΩ/√Hz BioZ sensitivity down to 1Hz. The ADC shows a 9.7b ENOB when sampled at 20ksps. The total power consumption of the AFE is 3.8μW.