IMPROVEMENT OF NEURAL NETWORK MODEL FOR PPG SIGNAL ANALYSIS
Keywords:
photoplethysmography (PPG), wavelet transformation, neural networks, machine learning, characteristic parametersAbstract
Purpose
This paper proposes to improve the method of predicting the age of the subject based on the recorded PPG signal.
Design/Methods/Approach
The PPG signal was obtained by the photoplethysmography method. The prediction method is based on the concept of a neural network, which uses wavelet transformation coefficients of the recorded PPG signal for training. The paper uses an RNN (Recurrent Neural Network) network, which was designed on five different architecture models. The focus of the work is on the comparison of the test results of these five different models and the selection of the model with the most acceptable results for further work. These five RNN models differ in the number of hidden layers. The first model has one hidden layer, the second has two hidden layers, etc., the fifth model has five hidden layers. A database with 500 recorded PPG signals is used to train the network.
Findings
The results of this paper open the discussion whether a larger number of hidden layers affects the prediction results and to what extent. There is also a discussion about whether to use RNN architecture or perhaps CNN (Convolutional Neural Network), which is again the topic of some future work in which a comparison of the results of these two architectures should be made.
Originality/Value
The use of wavelet transformation in the analysis of PPG signals can determine in a compromise way when certain changes occur in the signal and what their frequency content is. The goal of this work is in the domain of development, testing and addition of software solutions in the analysis of PPG signals, application of these solutions in artificial NN modeling and analysis of prediction results obtained by this method.
Keywords: photoplethysmography (PPG), wavelet transformation, neural networks, machine learning, characteristic parameters.
About the author
Zlatko Radovanović is a doctoral student at Alfa BK University - Faculty of Information Technologies. He graduated in electrical engineering in the field of automation. Areas of interest are signal analysis, system modeling and development and improvement of NN, AI.
Vojkan Nikolić, PhD is a Associate Professor on University of Criminal Investigation and Police Studies, Department of Information Technology, Belgrade. Areas of interest are information sistems, BI, data mining, NLP.
References
REFERENCES
Radovanović, Z., Jokić, S., Jokić, I., Gerazov, B., Kovačević, A. & Gligorić, N. (2025).
PPG signal analysis and wavelet selection for feature extraction. ALFATECH - Smart cities and
modern technologies. Proceedings (page 220). Belgrade: Alfa BK University.
DOI: 10.46793/ALFATECHproc25.220R
Allen, J. (2007). Photoplethysmography and its application in clinical physiological
measurement. Physiological Measurement, April 2007, 28(3).
DOI:10.1088/0967-3334/28/3/R01
Nikolic, V., Markoski, B., Kuk, K., Randjelovic, D., Cisar, P. (2017) . Modelling the System of
Receiving Quick Answers for e-Government Services: Study for the Crime Domain in the
Republic of Serbia. ACTA POLYTECHNICA HUNGARICA, 2017, vol. 14, No. 8, p. 143-163
DOI:10.12700/APH.14.8.2017.8.8
Popovic, S., Viduka, D., Basic, A., Dimic, V., Djukic, D., Nikolic, V., Stokic, A. (2025).
Optimization of Artificial Intelligence Algorithm Selection: PIPRECIA-S Model and Multi-
Criteria Analysis. ELECTRONICS, (2025), vol. 14, No. 3. DOI:10.3390/electronics14030562
Maksimovic, A., Nikolic, V., Vidojevic, D., Randjelovic, M., Djukanovic, S., Randjelovic, D.
(2024). Using Triple Modular Redundancy for Threshold Determination in DDOS Intrusion
Detection Systems. IEEE ACCESS, (2024), vol. 12. DOI:10.1109/ACCESS.2024.3384380
Stolic, P., Stevic, Z., Petronic, S., Nikolic, V., Stevic, M., Kreculj, D., Milosevic, D. (2024).
Modeling, Simulation, and Computer Control of a High-Frequency Wood Drying System.
ELECTRONICS, (2023), vol. 12, No. 1. DOI:10.3390/electronics12010226
Kevkic, T., Nikolic, V., Stojanovic, V., Milosavljevic, D., Jovanovic, S., (2022). Modeling
electrostatic potential in FDSOI MOSFETS: An approach based on homotopy perturbations.
OPEN PHYSICS, (2022), vol. 20, No. 1, p. 106-116. DOI:10.1515/phys-2022-0012
Raičević, S., Nikolić, V. (2024). Comparative analysis of certain clustering algorithms that do
not require a predefined number of clusters on the articles of the criminal code of the Republic
of Serbia. Journal of Computer and Forensic Sciences. 2024, vol. 3, No. 2, p. 28-42.
Milenković, J., Pavlović, M., Nikolić, V., Jasić, A., Premceski, V. (2020). Example of Clustering
Using K-Means Method in Python. International Conference on Information Technology and
Development of Education – ITRO 2020, Technical faculty "Mihajlo Pupin" Zrenjanin.
Janićijević, S., Nikolić, V. (2021). Graph Structures for Data Visualizations. Serbian Journal of
Engineering Management. 2021, Vol. 6, No. 2. DOI: 10.5937/SJEM2102024J
Avolio, A. (2002). The finger volume pulse and assessment of arterial properties. Journal of
Hypertension, 2002-12, Vol.20 (12), 2341-2343. DOI: 10.1097/00004872-200212000-00007
Webster, J., G. (2010). Medical Instrumentation: application and design. Wiley 2010.
Campolo, D. (2017). New Developments in Biomedical Engineering. InTech 2017.
Hertzman, A. (1938). The blood supply of various skin areas as estimated by the photoelectric
plethysmograph. American journal of physiology. 124 (2), 328-40.
Takazawa, K., T., N., Fujita, M., Matsuoka, O., Saiki, T., Aikawa, M., Tamura, S. & Ibukiyama
C., (1998). Assessment of vasocative agents and vascular aging by the second derivative of
photoplethysmogram waveform. Journal of Hypertension. 32, 65-70.
DOI: 10.1161/01.hyp.32.2.365
Rubins, U., Grabovskis, A., Grube, J. & Kukulis, I. (2008). Photoplethysmography Analysis of
Artery Properties in Patients with Cardiovascular Diseases. Springer Berlin Heidelberg.
Allen, J., & Kyriacou, P. (2021). Photoplethysmography - Technology, Signal Analysis and
Applications. eBook ISBN: 9780128235256.
Mallat, S. (1998). A wavelet Tour of Signal Processing. Academic Press, New York.
Radunović, D., P. (2005). Talasići. Akademska misao, Belgrade.
Stark, H., G. (2005). Wavelets and Signal Processing. University of Applied Sciences, Germany.
Springer Berlin Heidelberg.
INTERNET SOURCES
(URL1) Misiti, M., Misiti, Y., Oppenheim, G. & Poggi, J., M. (1997). Wavelet Toolbox For Use
with MATLAB, MathWorks. https://de.mathworks.com/help/wavelet/
(URL2) http://ataspinar.com/posts/machine-learning-with-signal-processing-techniques/
(URL3) Goodfellow, I., Bengio, Y. & Courvill, A. (2016). Deep Learning. e-book. MIT Press.
https://www.deeplearningbook.org/contents/ml.html
(URL4) https://github.com/PyWavelets/
(URL5) https://pywavelets.readthedocs.io/
(URL6) https://scikit-learn.org/
(URL7) https://github.com/stevanjokic
(URL8)
https://play.google.com/store/apps/details id=srb.ctb.pulse.heartrate.camera.ecg4everybody
(the application through which the PPG signals used in this work were collected and generated
in a file)
(URL9)
https://apxml.com/courses/getting-started-with-tensorflow/chapter-4-training-evaluating-
models/compiling-metrics
