Home » Blog » HRVCam: Robust Camera-Based Measurement of Heart rate Variability

HRVCam: Robust Camera-Based Measurement of Heart rate Variability

Published on 25 October 2021 at 22:01

Amruta Pai, Ashok Veeraraghavan, Ashutosh Sabharwal

Author Affiliations -Amruta Pai,1 Ashok Veeraraghavan,1 Ashutosh Sabharwal1
1Rice Univ. (United States)

Abstract.

Significance: Non-contact, camera-based heart rate variability estimation is desirable in numerous applications, including medical, automotive, and entertainment. Unfortunately, camera-based HRV accuracy and reliability suffer due to two challenges: (a) darker skin tones result in lower SNR and (b) relative motion induces measurement artifacts.

Aim: We propose an algorithm HRVCam that provides sufficient robustness to low SNR and motion-induced artifacts commonly present in imaging photoplethysmography (iPPG) signals.

Approach: HRVCam computes camera-based HRV from the instantaneous frequency of the iPPG signal. HRVCam uses automatic adaptive bandwidth filtering along with discrete energy separation to estimate the instantaneous frequency. The parameters of HRVCam use the observed characteristics of HRV and iPPG signals.

Results: We capture a new dataset containing 16 participants with diverse skin tones. We demonstrate that HRVCam reduces the error in camera-based HRV metrics significantly (more than 50% reduction) for videos with dark skin and face motion.

Conclusion: HRVCam can be used on top of iPPG estimation algorithms to provide robust HRV measurements making camera-based HRV practical.

Keywords: heart rate variability, imaging photoplethysmography, noncontact HRV, pulse frequency demodulation

 

1. Introduction

The nervous and the cardiac systems in the human body are intimately connected, primarily through the autonomous nervous system. This dynamic interplay is reflected in the beat-to-beat variation of the heart rate, formally labeled as heart rate variability (HRV). Interbeat interval (IBI) quantifies the time period between consecutive heartbeats. Several quantitative HRV metrics such as root mean square of successive differences in interbeat intervals (RMSSD) and standard deviation of interbeat intervals (SDNN) summarize the changes in the IBIs.,

HRV is clinically relevant because it provides a surrogate measure of the health of the autonomous nervous system. A low-baseline HRV is a symptom of poor autonomic function seen in diseases such as sudden cardiac death and diabetic autonomic neuropathy. Normal values of short term HRV metrics are 32 to 93 ms for SDNN and 19 to 75 ms for RMSSD.

HRV is clinically measured using electrocardiography (ECG) with well-defined controlled protocols. However, ECG can be limiting because the electrical leads need to be in contact with the skin surface. Contact may not always be feasible for applications such as driver stress detection, behavioral sensing, and monitoring for symptoms of sudden cardiac death in neonatal care units. Thus, many applications would benefit if robust camera-based HRV measurement were available.

Noncontact measurement of HRV may be possible with camera-based imaging photoplethysmography (iPPG), due to two factors. First, the optical photoplethysmography (PPG) signal enables the measurement of pulse rate variability. Pulse rate variability is shown to be correlated to HRV. Second, the PPG signal can be captured by the camera placed at a distance from the participant.

Noncontact HRV estimation suffers from several disadvantages compared to contact HRV estimation. First, camera-based methods result in low SNR due to the absorption of incident light by high amounts of melanin pigment in dark skin tissue. Second, camera-based noncontact methods have to contend with unpredictable illumination changes due to relative nonrigid movements of the skin surface. The unpredictable illumination changes corrupt the shape of the iPPG signal rendering crucial IBIs not easily measurable. The disadvantages are prominent in iPPG signals because both the light source (e.g., ambient light) and light detector (i.e., the camera) are at a distance from the skin surface.

A standard time-based method to measure HRV is to detect peaks in the PPG or ECG signal and then estimate HRV from the measured time differences of the occurrence of the peaks. However, peak-based approaches typically perform poorly due to the low SNR and often high-motion-related artifacts in iPPG signals.

An alternate approach to peak-based estimation is to measure HRV using pulse frequency demodulation (PFDM), that relies on the instantaneous frequency. Chou et al. demonstrated that the frequency demodulation approach was more robust than the peak selection method for noisy contact photoplethysmography (cPPG) signals.

We investigated the use of PFDM to improve the accuracy of HRV metrics measured from a low signal quality iPPG signal. The main contributions of this paper are twofold:

  •  HRVCam algorithm. We propose HRVCam, a new algorithm based on a frequency demodulation framework to estimate the instantaneous frequency of the iPPG signal. The framework is a combination of a new automated adaptive bandpass filter and the discrete energy separation algorithm (DESA).
  •  HRVCam dataset. We collected a new iPPG dataset with validated ground truth using a pulse oximeter under different scenarios: (i) low melanin pigment (light skin tones), (ii) high melanin pigment (dark skin tones), (iii) low motion, such as sitting still, and (iv) different degrees of natural motion (reading, watching, and talking). The new dataset is publicly available. Evaluation of HRVCam on the dataset shows improved performance of HRVCam when compared to existing state-of-the-art approaches.

More 

 


«   »

Add comment

Comments

There are no comments yet.