• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.2
• /
• pp.51-58
• /
• 2023

This paper proposes a deep learning method for estimating the heart rate from facial videos. Our proposed method estimates remote photoplethysmography (rPPG) signals to predict the heart rate. Although there have been proposed several methods for estimating rPPG signals, most previous methods can not be utilized in low-power single board computers due to their computational complexity. To address this problem, we construct a lightweight student model and employ a knowledge distillation technique to reduce the performance degradation of a deeper network model. The teacher model consists of 795k parameters, whereas the student model only contains 24k parameters, and therefore, the inference time was reduced with the factor of 10. By distilling the knowledge of the intermediate feature maps of the teacher model, we improved the accuracy of the student model for estimating the heart rate. Experiments were conducted on the UBFC-rPPG dataset to demonstrate the effectiveness of the proposed method. Moreover, we collected our own dataset to verify the accuracy and processing time of the proposed method on a real-world dataset. Experimental results on a NVIDIA Jetson Nano board demonstrate that our proposed method can infer the heart rate in real time with the mean absolute error of 2.5183 bpm.

• Proceedings of the Korean Magnestics Society Conference
• /
• 2013.05a
• /
• pp.38-40
• /
• 2013

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.41 no.5
• /
• pp.65-70
• /
• 2004

In this paper, we present a new methodology to analyze the characteristic of artery by using 4 channels photoplethysmograpy. The proposed parameter is a time difference of pulse transit time(PTT) between Pulse waves at finger site and at toe site. To verify the usefulness of the developed system volume pulse waves on 4 sites were measured simultaneously for total 51 normal subjects (male 26 and female 25) aged from 9 to 83 years old. And then correlations between the analysis parameters and age were evaluated by using linear regression analysis method. As the result of experiments, the change of parameter was found according to ages. The result of regression analysis about relationships between the parameter and ages for n=51, the coefficient of correlation of non-normalized data has 0.79770 in left side and 0.80599 in right side and the coefficient of correlation of normalized data by height has 0.81345 in left side and 0.81605 in right side.

• Journal of Korea Society of Industrial Information Systems
• /
• v.25 no.3
• /
• pp.1-10
• /
• 2020

In this paper, we propose an algorithm for estimating blood pressure using ECG (Electrocardiogram) and PPG (Photoplethysmography) signals. To estimate the BP (Blood pressure), we generate a periodic input signal, remove the noise according to the differential and threshold methods, and then estimate the systolic and diastolic blood pressures based on the convolutional neural network. We used 49 patient data of 3.1GB in the MIMIC database. As a result, it was found that the prediction error (RMSE) of systolic BP was 5.80mmHg, and the prediction error of diastolic BP was 2.78mmHg. This result confirms that the performance of class A is satisfied with the existing BP monitor evaluation method proposed by the British High Blood Pressure Association.

• The Journal of Korean Medicine
• /
• v.29 no.2
• /
• pp.60-70
• /
• 2008

Objectives: The aim of the present investigation was to determine whether the finger photoplethysmography (PPG) of the young will become analogous to those of the old when the PPG harmonic components in the young decrease. Methods: The PPG was measured in 46 old men (21 males and 25 females) over the age of 60 years and 10 young men (5 males and 5 females). We acquired the representative pulse waveform of old men by averaging the PPG waveforms measured in the old men. after the PPG harmonic components in the young men were diminished with notch filtering, we compared the representative pulse waveform of old men and those of the young men. Results: The PPG waveform of the young males became analogous to the representative pulse waveform of the old as the PPG harmonic components in the young were diminished with notch filtering. Especially after the second harmonic component in the young was diminished with notch filtering, increase of the Pearson's correlation coefficient was prominent. On the other hand, the Pearson's correlation coefficient between the PPG waveforms of the young females and that of the old deceased after the second harmonic component diminished in the young. Conclusions: Decrease of the second harmonic component may be a precondition of typical age-related change of the pulse waveform in the healthy male.

• Journal of Biomedical Engineering Research
• /
• v.34 no.3
• /
• pp.105-110
• /
• 2013

Respiratory signal is one of the important physiological information indicating the status and function of the body. Recent studies have provided the possibility of being able to estimate the respiratory signal by using a change of PWV(pulse width variability), PRV(pulse rate variability) and PAV(pulse amplitude variability) in the PPG (photoplethysmography) signal during daily life. But, it is not clear whether the respiratory monitoring is possible even during sleep. Therefore, in this paper, we estimated the respiration from PWV, PRV and PAV of PPG signals during sleep. In addition, respiration rates of the estimated respiration signal were calculated through a time-frequency analysis, and errors between respiration rates calculated from each parameter and from reference signal were evaluated in terms of 1 sec, 10 sec and 1 min. As a result, it showed the errors in PWV(1s: $36.38{\pm}37.69$ mHz, 10s: $36.53{\pm}38.16$ mHz, 60s: $30.35{\pm}38.72$ mHz), in PRV(1s: $1.45{\pm}1.38$ mHz, 10s: $1.44{\pm}1.37$ mHz, 60s: $0.45{\pm}0.56$ mHz), and in PAV(1s: $1.05{\pm}0.81$ mHz, 10s: $1.05{\pm}0.79$ mHz, 60s: $0.56{\pm}0.93$ mHz). The errors in PRV and PAV are lower than that of PWV. Finally, we concluded that PRV and PAV are more effective than PWV in monitoring the respiration in daily life as well as during sleep.

• Journal of the Ergonomics Society of Korea
• /
• v.29 no.1
• /
• pp.113-120
• /
• 2010

This research is to evaluate the arousal level by using cardiovascular response. PPG was used in this study as one of the method of measuring it rather than ECG (Electrocardiography) for the purpose of solving ergonomic problem of sensing. The participants were in the age group of 20 (mean=24, standard deviation=1.25): five men and five women. Each experiment composed with four identical sets. First, a black screen was displayed for 30 second rest. Then, the prepared 6 pair images were randomly presented for 10 second stimulation and for 30 second non-stimulation. PPG was measured on the earlobes of experimenters at 200Hz sampling frequency. PPG amplitude, PPI(Pulse to Pulse Interval), and PRV(Pulse Rate Variability) were analyzed according to arousal level. T-test was performed to compare between the PPG variables of rest and relaxation, rest and arousal, and relaxation and arousal. Relative to the rest state, PPG amplitude decreased in relaxed state and increased in aroused state. Relative to the rest state, PPI decreased in both emotional states. However, more significant decline was observed in aroused state. PRV's LF and HF were used in the form of LF/HF to compare between the relaxed and the aroused state. Therefore, PPG signal showed significant differences between relaxed and aroused state. In conclusion, evaluation of human arousal level used in the PPG analysis demonstrated that PPG has better usability and comforter measurement than ECG and is clearly an alternative method of measuring arousal level.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.8
• /
• pp.18-26
• /
• 2015

Measuring the heart rate during exercise is important to properly control the amount of exercise. With the recent advent of smart device usage, there is a dramatic increase in interest in devices for the real-time measurement of the heart rate during exercise. During intensive exercise, accurate heart rate estimation from wrist-type photoplethysmography (PPG) signals is a very difficult problem due to motion artifact (MA). In this study, we propose an efficient algorithm for an accurate estimation of the heart rate from wrist-type PPG signals. For the twelve data sets, the proposed algorithm achieves the average absolute error of 1.38 beat per minute (BPM) and the Pearson correlation between the estimates and the ground-truth of heart rate was 0.9922. The proposed algorithm presents the strengths in an accurate estimation together with a fast computation speed, which is attractive in application to wearable devices.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.8
• /
• pp.1431-1437
• /
• 2016

In mobile healthcare device, when to measure the heart rate by using the PPG signal, its performance is reduced according to the motion artifacts that is the movement of user. This is because the frequency range of motion (0.01-10 Hz) and that of PPG signals overlap. Also, the motion artifacts cannot be rectified by general filters. To solve the problem, this paper proposes a method using filter banks and independent component analysis (ICA). To evaluate the performance of the proposed method, we were artificially applied various movements and compared heart rate errors of the moving average filter and ICA. In the experimental results, heart rate error of the proposed method showed very low than moving average filter and ICA. In this way, it is possible to measure stable heart rate if the proposed method is applied to the healthcare terminal design.

• Journal of the Institute of Convergence Signal Processing
• /
• v.20 no.2
• /
• pp.91-98
• /
• 2019

Accidents of worker that occur in isolated places are difficult to rescue, unlike general construction accidents. There is a problem of communication limitation when an accident occurs in an isolated place. Also, it is difficult to search the accident place due to the absence of CCTV. In order to solve these problems, this paper proposes a device that combines IoT technology with a safety helmet, which must be worn in the workplace. The proposed device additionally designs and implements a real-time PPG(Photoplethysmography) sensor, body temperature sensor, accelerometer sensor and a camera sensor on the helmet. The proposed helmet system allows the user and the control center to monitor the state of the worker. In addition, when an abnormal biological signal or fall occurs to the worker, the image is transmitted to the control center. By using the proposed system, it is possible to check the status of the worker in real time, so that it has an advantage that it can cope with the accident quickly.