SEQUENTIAL NOISE SUPPRESSION ALGORITHM FROM THE SPEECH SIGNAL

Abstract

The problem of denoising speech signals is a relevant task for speech recognition systems, mobile communication devices, and voice interfaces. This study proposes a new approach that combines Self - Organizing Map (SOM) and Spectral Subtraction methods. This approach effectively denoises speech by using the SOM neural network to detect noisy clusters and combining energy and frequency features. During the study, the Minimum Statistics Noise Estimation method was used for noise estimation, an adaptive feature selection strategy was applied, and experiments were conducted at various noise levels (from 1% to 25%). The results, evaluated using the PESQ metric, showed that the proposed approach outperforms traditional methods such as Wavelet and Spectral Subtraction. The advantage of this approach lies in the improved algorithm for detecting noisy clusters and the optimized post-processing stage aimed at preserving speech naturalness. The research results indicate that the combination of the SOM and Spectral Subtraction methods is an effective solution for denoising speech signals in noisy environments.

Keywords

References

1. Mamatov, N., Niyozmatova, N., & Samijonov, A. (2021). Software for preprocessing voice signals. International Journal of Applied Science and Engineering, 18(1), 1 - 8. INNOVA TSI ON TEXNOLOGIYALAR INNOVATIVE TECHNOLOGIES ИННОВАЦИОННЫЕ ТЕХНОЛОГИИ 2025 - yil 1(58) - son 2025 volume 58, number 2 Том 58 No 2, 2025 ISSN 2181 - 4732 ISSN 2181 - 4732 129
2. Niyozmatova, N. N., Jalelov, N. K., Samijonov, N. B., & Madrahimova, N. M. (2024). Eliminating noise from a speech signal based on a pair of filters. International Journal of Science and Research Archive, 13(2), 401 – 410. https://doi.org/10.30574/ijsra.2024.13.2.2058
3. Boll, S.F. (1979). Suppression of acoustic noise in speech using spectral subtraction. IEEE Transactions on Acoustics, Speech, and Signal Processing, 27(2), 113 - 120.
4. Lim, J.S., & Oppenheim, A.V. (1979). Enhancement and bandwidth compression of noisy speech. Proceedings of the IEEE, 67(12), 1586 - 1604.
5. Ephraim, Y., & Malah, D. (1984). Speech enhancement using a minimum - mean square error short - time spectral amplitude estimator. IEEE Transactions on Acoustics, Speech, and Signal Processing, 32(6), 1109 - 1121.
6. Ramirez, J., Górriz, J.M., & Segura, J.C. (2007). Voice activity detection: fundamentals and speech recognition system robustness. In M. Grimm & K. Kroschel (Eds.), Robust Speech Recognition and Understanding (pp. 1 - 22). I - Tech Education and Publishing.
7. Martin, R. (2001). Noise power spectral density estimation based on optimal smoothing and minimum statistics. IEEE Transactions on Speech and Audio Processing, 9(5), 504 - 512.
8. Kohonen, T. (1982). Self - organized formation of topologically correct feature maps. Biological Cybernetics, 43(1), 59 - 69.
9. Zhang, X.L., & Wu, J. (2013). Deep belief networks based voice activity detection. IEEE Transactions on Audio, Speech, and Language Processing, 21(4), 697 - 710.
10. Ramírez, J., Segura, J.C., Benítez, C., De La Torre, A., & Rubio, A. (2004). Efficient voice activity detection algorithms using long - term speech information. Speech Communication, 42(3 - 4), 271 - 287.