HeartBeatNet: Enhancing Fast and Accurate Heart Rate Estimation With FMCW Radar and Lightweight Deep Learning
Dongryul Kim, Jaeyoung Choi, J. N. Yoon, Sungpil Cheon, Byungkwan Kim
Abstract
In this letter, we propose an approach for accurately estima-ting heart rate in a short time using frequency-modulated continuo- us-wave (FMCW) radar and a lightweight deep learning method. Radar signals obtained from measuring a human are preprocessed using unique signal processing techniques to represent the characteristics of respiration and heart rates, enabling effective network training. In addition, we introduce the lightweight deep learning network, i.e., HeartBeatNet, to directly estimate heart rate with high accuracy and low computational cost using skip connections and dilated convolution. The experimental results demonstrate that our proposed approach can accurately estimate heart rate in a short time and show potential for real-world industrial applications due to its lightweight design.