1. ISTCMTM
  2. All Volumes and Issues
  3. Volume 86, no.3, 2025
  4. IMPACT OF AUDIO SIGNAL DURATION ON THE ACCURACY OF SPEAKER VOICE IDENTIFICATION

IMPACT OF AUDIO SIGNAL DURATION ON THE ACCURACY OF SPEAKER VOICE IDENTIFICATION

ISTCMTM.
2025;
Volume 86(3), Number 3
: 5-9
https://doi.org/10.23939/istcmtm2025.03
Authors:
  1. Volodymyr Povoroznyk
  2. Ihor Mykytyn
1
Lviv Polytechnic National University
2
Lviv Polytechnic National University

This paper investigates the capability of a system based on voice embeddings to identify speakers. We use a set of audio recordings from five speakers and construct clips of varying durations – 5 to 600 seconds. Pyannote-audio embeddings are extracted by a neural network, after which similarity coefficients are computed between embeddings of clips from the same speaker (intra-speaker similarity) and from different speakers (inter-speaker dissimilarity). We study how clip duration affects the protection zone when separating speakers into “own/other.” The experiments show that there exists a certain clip duration that yields a relatively wide protection zone, which raises the probability of accurate voice-based identification. The results may be used in future research on biometric verification

Voice biometrics
voice embedding
neural networks
speaker identification
similarity coefficient 1. Introduction
  1. Desplanques, B., Thienpondt, J., & Demuynck, K. (2020). ECAPA-TDNN: Emphasized Channel Attention, Propa- gation and Aggregation in TDNN Based Speaker Verifi- cation.              Interspeech  2020.       [Online]. Available: https://www.isca-speech.org/archive/Interspeech_2020/ pdfs/1137.pdf
  2. Bredin, H., Laurent, A., & Gillies, A. (2020). Pyannote.audio: neural building blocks for speaker diarization. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2020). [Online]. Available: https://ieeexplore.ieee.org/document/9053198
  3. Ruda, H., Sabodashko, D., Mykytyn, H., Shved, M., Borduliak, S., & Korshun, N. (2024). Specifics of creating and distributing phishing web resources. Cybersecurity: Education, Science, Technique, 2(12), 80–88. Available: https://csecurity.kubg.edu.ua/index.php/journal/article/view/ 645/508
  4. A. Levy, B. Riva Shalom, and M. Chalamish, “A guide to similarity measures and their data science applications,” Journal of Big Data, vol. 12, Art. no. 188, 2025. [Online]. Available:    https://journalofbigdata.springeropen.com/ articles/10.1186/s40537-025-01227-1
  5. Google Cloud, “Chirp 3 HD: High-quality, low-latency text- to-speech model,” Google Cloud Text-to-Speech, 2024. [Online]. Available: https://cloud.google.com/text-to- speech/docs/chirp3-hd
  6. B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA- TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,” arXiv preprint arXiv:1911.01255, 2019. [Online]. Available: https://arxiv.org/abs/1911.01255
  7. H. Bredin, A. Laurent, and Y. Zhong, “End-to-end domain- adversarial voice activity detection,” Proc. Interspeech 2021, pp. 4658–4662, 2021. [Online]. Available: https:// www.isca-archive.org/interspeech_2021/bredin21_ interspeech.pdf
  8. E. F. Krause, Taxicab Geometry: An Adventure in Non- Euclidean Geometry. New York, NY, USA: Dover Publications, 1986.
  9. T. Souravlas, I. Roumeliotis, C. Roumeliotis, and C. Zissis, “Time series similarity measures and deep learning: State-of- the-art review,” arXiv preprint arXiv:2412.20574, 2024. [Online]. Available: https://arxiv.org/abs/2412.20574