Motion- Based Respiratory Rate Estimation with Motion Artifact Removal Using Video of Face and Upper Body

Gwak, Migyeong; Vatanparvar, Korosh; Kuang, Jilong; Gao, Alex

doi:10.1109/embc48229.2022.9871231

Cited by 13 publications

(12 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…5), we compare CalibrationPhys with state-of-the-art methods for HR measurement on the smartphone. As comparative methods, we used three signal processing methods 7 (ICA [1], CHROM [3], and POS [4]), four supervised DL-based methods 8 (DeepPhys [12], Phys-Net [15], TS-CAN [17], and EfficientPhys-C [19]), and one self-supervised DL-based method (Contrast-Phys [27]). For the signal processing methods, a single ROI was utilized as input by averaging the RGB signals for the same ROI as the proposed method.…”

Section: Resultsmentioning

confidence: 99%

“…For RR measurement, the chest and head movements associated with respiratory motion, i.e., respiratory waves, are extracted. RR estimation methods obtain pixel movement based on optical flow [5]- [7] or pixel intensity variation [8]- [10] from chest or facial ROIs. Zhan et al [11] showed that pixel movement-based methods perform better RR estimation than pixel intensity variationbased methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Heart Rate and Oxygen Saturation Estimation from Facial Video with Multimodal Physiological Data Generation

Akamatsu

Onishi

Imaoka

2022

ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Video-based heart and respiratory rate measurements using facial videos are more useful and user-friendly than traditional contact-based sensors. However, most of the current deep learning approaches require ground-truth pulse and respiratory waves for model training, which are expensive to collect. In this paper, we propose CalibrationPhys, a self-supervised video-based heart and respiratory rate measurement method that calibrates between multiple cameras. CalibrationPhys trains deep learning models without supervised labels by using facial videos captured simultaneously by multiple cameras. Contrastive learning is performed so that the pulse and respiratory waves predicted from the synchronized videos using multiple cameras are positive and those from different videos are negative. CalibrationPhys also improves the robustness of the models by means of a data augmentation technique and successfully leverages a pre-trained model for a particular camera. Experimental results utilizing two datasets demonstrate that CalibrationPhys outperforms stateof-the-art heart and respiratory rate measurement methods. Since we optimize camera-specific models using only videos from multiple cameras, our approach makes it easy to use arbitrary cameras for heart and respiratory rate measurements.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Heart Rate and Oxygen Saturation Estimation from Facial Video with Multimodal Physiological Data Generation

Akamatsu

Onishi

Imaoka

2022

ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…Also, some studies present techniques to exclude motion artifacts in the respiratory signal. For example, [46] used an IMU sensor to clean a respiratory signal collected with piezoresistive sensors by implementing an algorithm in a frequency domain analysis. Furthermore, [47] proposed a method for filtering motion artifacts on a respiratory signal extracted from a PPG signal using information provided by an IMU sensor.…”

Section: Discussionmentioning

confidence: 99%

A Signal Quality Index for Improving the Estimation of Breath-by-Breath Respiratory Rate During Sport and Exercise

Romano,

Innocenti,

Schena

et al. 2023

IEEE Sensors J.

View full text Add to dashboard Cite

The breath-by-breath measurement of respiratory rate (RR) plays a pivotal role in sports and exercise. The accurate estimation of RR values on a breath-by-breath basis with wearable sensors has several open challenges during training, including motion artefacts and other breathingunrelated events. This paper presents a novel method based on a signal quality index (SQI) for identifying and excluding unreliable breaths from breathing waveforms. The method analyses the morphological characteristics of the respiratory signal, comparing each breath with an average breath template calculated as an average of all individual breaths. The comparison is made using a template matching without the need of a reference signal. Experimental tests have been carried out at rest and during walking, running, and cycling activities to assess the method's performance in estimating breath-by-breath RR by comparison with reference values collected with a flowmeter. The comparison between the RR values has been performed with an ad hoc developed method able to accomplish this task, even when the number of breaths identified by the two devices is different. The obtained results showed that our SQI-based method improves the accuracy of RR estimation by reducing the Mean Absolute Percentage Error (MAPE) values in all the tested conditions (18.5%, 22.2%, 2.8% and 14.1% of MAPE improvement rate during rest, walking, running and cycling, respectively). Pilot tests during high-intensity interval training (HIIT) also demonstrated a 30.7% MAPE improvement rate. The promising findings demonstrated that using SQI-based algorithms can lead to more accurate RR estimations during exercise by using comfortable wearable sensors.

show abstract

“…The respiratory rate denotes the count of respiratory cycles (breaths) occurring within a specific time frame, constituting a fundamental physiological function in humans [2]. The anticipation and prompt identification of health concerns and respiratory disorders hinge upon the analysis of respiratory signals and rates [1,[3][4][5][6][7][8][9][10][11]. Moreover, specific respiratory conditions necessitate meticulous respiratory system supervision along with physical health training [12].…”

Section: Introductionmentioning

confidence: 99%

“…However, despite their high accuracy, contactbased respiratory-rate measurements are not recommended for those with sensitive skin, preterm newborns, or the elderly because they obstruct normal breathing and are uncomfortable for patients who need to wear wearables for long-term monitoring. Sharing infected devices in institutions such as hospitals and nursing homes may also lead to the spread of illnesses [10,11]. Studies have been conducted to develop noncontact and non-invasive methods for acquiring respiratory signals/rates to address these issues.…”

Section: Introductionmentioning

confidence: 99%

Facial Video‐Based Robust Measurement of Respiratory Rates in Various Environmental Conditions

Park,

Hong

2023

Journal of Sensors

View full text Add to dashboard Cite

This paper addresses the process of respiration, which involves providing oxygen to the body’s cells, blood, and tissues. The significance of respiratory signals as predictors of diverse physical conditions, including respiratory health and disorders, is emphasized. Technologies such as pressure-sensor belts for chest or abdominal movement detection, carbon-dioxide sensors placed under the nose, and photoplethysmography (PPG) have been developed for real-time monitoring of respiratory signals. However, wearable devices with direct body contact may inconvenience sleeping patients, newborns, or individuals with sensitive skin. Noncontact and noninvasive respiratory rate measurements utilizing facial videos have been explored, albeit within controlled environments. This study introduces a novel approach to measure robust respiratory signals and facial respiratory rates (FRR) using facial videos. The proposed method employs bounding box stabilization to eliminate high-frequency components stemming from face detection. Real-time coordinate correction is implemented to ensure accurate respiration signal calculation across varying environments. The method defines a region of interest (ROI) on the face, computes RGB data, and transforms the RGB channel into a YCgCo channel. The respiratory signal is subsequently derived by applying partial zero padding-based fast Fourier transform (FFT) and inverse FFT (iFFT) to the Cg signal within the frequency band associated with respiration. The FRR is accurately measured by analyzing peak-to-peak intervals within the respiratory signal. The efficacy of the proposed method is demonstrated through a comparative analysis of respiratory rates calculated from PPG devices and facial videos across four scenarios (indoor, outdoor, car, and drone). The results highlight the reliability and precision of the suggested approach in accurately computing respiratory rates across diverse environments.

show abstract

Motion- Based Respiratory Rate Estimation with Motion Artifact Removal Using Video of Face and Upper Body

Cited by 13 publications

References 24 publications

Heart Rate and Oxygen Saturation Estimation from Facial Video with Multimodal Physiological Data Generation

Heart Rate and Oxygen Saturation Estimation from Facial Video with Multimodal Physiological Data Generation

A Signal Quality Index for Improving the Estimation of Breath-by-Breath Respiratory Rate During Sport and Exercise

Facial Video‐Based Robust Measurement of Respiratory Rates in Various Environmental Conditions

Contact Info

Product

Resources

About