A 10.60 $\mu$W 150 GOPS Mixed-Bit-Width Sparse CNN Accelerator for Life-Threatening Ventricular Arrhythmia Detection

TL;DR

This paper introduces an ultra-low power, sparse mixed-bit-width CNN accelerator that efficiently detects ventricular arrhythmia with high accuracy, suitable for implantable medical devices, achieving significant power and size reductions.

Contribution

The paper presents a novel sparse, mixed-bit-width CNN accelerator with high efficiency and low power consumption, optimized for medical device applications.

Findings

Achieves 10.60 μW power consumption and 150 GOPS performance.

Demonstrates 50% sparsity in a quantized 1D CNN.

Power density is 14.23 times lower than state-of-the-art solutions.

Abstract

This paper proposes an ultra-low power, mixed-bit-width sparse convolutional neural network (CNN) accelerator to accelerate ventricular arrhythmia (VA) detection. The chip achieves 50% sparsity in a quantized 1D CNN using a sparse processing element (SPE) architecture. Measurement on the prototype chip TSMC 40nm CMOS low-power (LP) process for the VA classification task demonstrates that it consumes 10.60 $\mu$W of power while achieving a performance of 150 GOPS and a diagnostic accuracy of 99.95%. The computation power density is only 0.57 $\mu$W/mm$^2$, which is 14.23X smaller than state-of-the-art works, making it highly suitable for implantable and wearable medical devices.