Internship Report: Benchmark of Deep Learning-based Imaging PPG in Automotive Domain

TL;DR

A benchmark of a NIR-based iPPG method using a deep learning model is provided by evaluating its performance on MR-NIRP Car dataset and suggesting that while the method shows promise, further improvements are needed to make it reliable for real-world driving conditions.

Abstract

Imaging photoplethysmography (iPPG) can be used for heart rate monitoring during driving, which is expected to reduce traffic accidents by continuously assessing drivers' physical condition. Deep learning-based iPPG methods using near-infrared (NIR) cameras have recently gained attention as a promising approach. To help understand the challenges in applying iPPG in automotive, we provide a benchmark of a NIR-based method using a deep learning model by evaluating its performance on MR-NIRP Car dataset. Experiment results show that the average mean absolute error (MAE) is 7.5 bpm and 16.6 bpm under drivers' heads keeping still or having small motion, respectively. These findings suggest that while the method shows promise, further improvements are needed to make it reliable for real-world driving conditions.

Figures (6)

• Figure 1: Graphical representation of overall procedure of the deep learning-based NIR iPPG method proposed in 9506663
• Figure 2: U-net model-based Deep learning module used for pulse waveform estimation. Input for the model is 300-length time series with 23 dimensions and it outputs a 300-length, single-channel time series.
• Figure 3: Visualization examples of heart rate estimation
• Figure 4: Visualization examples of the spectrogram of estimated pulse waveform
• Figure 5: Subject-wise performance comparison demonstrating impact of data augmentation on Still condition of Garage scenario. DA stands for data augmentation