A.K. Alimuradov1, B.A. Porezanov2, A.Yu. Tychkov3, D.S. Chernyshov4, M.A. Evdokimova5, M.R. Gamidullaeva6, K.M. Gadzhiev7, M.Ya. Karibov8

1,3,4,5,7 Penza State University (Penza, Russia)

2 JSC "CeSIS" (Penza, Russia)

6 First Moscow State Medical University n.a. I.M. Sechenov (Sechenov University) (Moscow, Russia)

8 Dagestan State Technical University (Makhachkala, Russia)

1 alansapfir@yandex.ru, 2 bogdan.porezanov@yandex.ru, 3 tychkov-a@mail.ru, 4 chernyshov.ds@mail.ru, 5 m.evdok007@gmail.com, 6 madi2707@yandex.ru, 7 gadzhievkkk@gmail.com, 8 dstu-u831@mail.ru

Segmentation of a speech signal into "speech/pause" segments is a critical preprocessing step that determines the accuracy of subsequent algorithms in fields such as speech recognition, speaker identification, and biomedical voice analysis. Existing frequency and time domain segmentation methods often demonstrate insufficient accuracy under conditions of natural speech variability and background noise, which limits their application in medical diagnostics, especially when analyzing voice disorders.

Aim – improving frequency and time domain segmentation methods based on various modifications of empirical mode decomposition. The essence of the improvement is to transform the original speech signal into a set of mode signals using classical, ensemble, and improved decomposition for subsequent "speech/pause" segmentation.

The paper presents a comparison of the original speech signal and signals reconstructed from selected mode components obtained using various modifications of empirical mode decomposition. For all decomposition methods, the best result was provided by a combination of the first, second, and fourth mode signals. According to the research results, all decomposition methods improved segmentation accuracy compared to the original signal. The best result is achieved with improved complete ensemble empirical mode decomposition with adaptive noise – the type I and type II errors were 0.93% and 1.76%, respectively, and their sum was 2.69% compared to 5.92% for the original signal.

The developed improved segmentation methods can be effectively applied in biomedical voice analysis systems. In particular, high accuracy in isolating speech fragments is critically important for the objective assessment of voice disorders resulting from complications after surgical interventions on the neck organs.

Alimuradov A.K., Porezanov B.A., Tychkov A.Yu., Chernyshov D.S., Evdokimova M.A., Gamidullaeva M.R., Gadzhiev K.M., Karibov M.Ya. Improving frequency and time domain speech signal segmentation methods based on EMD // Biomedical Radioelectronics. 2026. V. 29. № 6. P. 63−71. DOI: https://doi.org/10.18127/j15604136-202606-12

1. Tseng C.-Y., Chang C.-H. Pause or No Pause? Prosodic Phrase Boundaries Revisited. Tsinghua Science and Technology. 2008. V. 13. № 4. P. 500–509.
2. Prasetio B.H., Widasari E.R., Tamura H. Automatic Multiscale-based Peak Detection on Short Time Energy and Spectral Centroid Feature Extraction for Conversational Speech Segmentation. Proceedings of the 2021 International Conference on Sustainable Information Engineering and Technology. New York: ACM, 2021. P. 44–49. DOI: 10.1145/3479645.3479675.
3. Endah S.N., Fadlilah N., Kusumaningrum R., Adhy S. Continuous Speech Segmentation Using Dynamic Thresholding of Short-Term Features. Journal of Engineering Science and Technology. 2022. V. 17. № 4. P. 2919–2935.
4. Ait Mait H., Aboutabit N. Unsupervised VAD Method Based on Short-Time Energy and Spectral Centroid in Arabic Speech Case. International Journal of Computer Applications in Technology. 2024. V. 74. № 3. P. 158–170. DOI: 10.1504/IJCAT.2024.141931.
5. Alimuradov A.K., Tychkov A.Yu., CHurakov P.P., Dudnikov D.S. Novyj podhod segmentacii rechevyh signalov na osnove chastotno-vremennogo analiza. Izmerenie. Monitoring. Upravlenie. Kontrol'. 2022. № 4. S. 80–92. DOI: 10.21685/2307-5538-2022-4-11 (In Russian).
6. Alimuradov A.K., Tychkov A.Yu., Simakova O.S., Mamonova A.A., Yuldashev Z.M., Temirova D.A. Tekhnologiya segmentacii «signal/pauza» na osnove analiza urovnya smeshivaniya fragmentov rechevyh signalov. Biomedicinskaya radioelektronika. 2025. T. 28, № 2. S. 38–43 DOI: https://doi.org/10.18127/j15604136-202502-06 (In Russian).
7. Alimuradov A.K. Povyshenie effektivnosti segmentacii rechevyh signalov na osnove energeticheskogo operatora Tigera. Izmereniya. Monitoring. Upravlenie. Kontrol'. 2021. № 3. S. 80–92. DOI: 10.21685/2307-5538-2021-3-10 (In Russian).
8. Kaiser J.F. On a Simple Algorithm to Calculate the Energy of a Signal. 1990 International Conference on Acoustics, Speech, and Signal Processing. Piscataway: IEEE, 1990. V. 1. P. 381–384. DOI: 10.1109/ICASSP.1990.115702.
9. Huang N.E., Shen Z., Long S.R., Wu M.C., Shih H.H., Zheng Q., Yen N.-C., Tung C.C., Liu H.H. The Empirical Mode Decomposition and the Hilbert Spectrum for Nonlinear and Non-stationary Time Series Analysis. Proceedings of the Royal Society of London. Series A. 1998. V. 454, № 1971. P. 903–995.
10. Wu Z., Huang N.E. Ensemble Empirical Mode Decomposition: A Noise-Assisted Data Analysis Method. Advances in Adaptive Data Analysis. 2009. V. 1. № 1. P. 1–41.
11. Colominas M.A., Schlotthauer G., Torres M.E. Improved Complete Ensemble EMD: A Suitable Tool for Biomedical Signal Processing. Biomedical Signal Processing and Control. 2014. V. 14. P. 19–29. DOI: 10.1016/j.bspc.2014.06.009.