izmertechИзмерительная техникаIzmeritel`naya Tekhnika0368-10252949-5237ФГУП "ВНИИФТРИ"10.32446/0368-1025it.2024-5-54-63izmertech-2139Research ArticleАКУСТИЧЕСКИЕ ИЗМЕРЕНИЯACOUSTIC MEASUREMENTSМетод тестирования устойчивости авторегрессионной модели речевого тракта и корректировки её параметровMethod for testing the stability of an autoregressive model of the vocal tract and adjusting its parametershttps://orcid.org/0000-0003-3045-3337СавченкоВ. В.SavchenkoV. V.

Владимир Васильевич Савченко, доктор технических наук, профессор

Нижний Новгород

Vladimir V. Savchenko

Nizhny Novgorod

vvsavchenko@yandex.ruhttps://orcid.org/0000-0002-2776-5471СавченкоЛ. В.SavchenkoL. V.

Людмила Васильевна Савченко

Нижний Новгород

Lyudmila V. Savchenko

Nizhny Novgorod

avsavchenko@hse.ruНациональный исследовательский университет «Высшая школа экономики»РоссияNational Research University Higher School of EconomicsRussian Federation202410062024055463Copyright © Савченко В.В., Савченко Л.В., 20242024Савченко В.В., Савченко Л.В.Savchenko V.V., Savchenko L.V.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://www.izmt.ru/jour/article/view/2139

В рамках традиционного направления исследований в области акустических измерений рассмотрена авторегрессионная модель речевого тракта как ключевого звена речевого аппарата человека. Указано на острую проблему обеспечения устойчивости авторегрессионной модели в системах с адаптацией параметров под наблюдаемый речевой сигнал небольшой длительности. Для преодоления указанной проблемы поставлена задача тестирования устойчивости авторегрессионной модели и корректировки её параметров по результатам тестирования. В основу исследования положена авторская методика формантного анализа гласных звуков речи через синтез рекурсивного формирующего фильтра в режиме свободных колебаний. Для решения поставленной задачи предложен метод тестирования устойчивости авторегрессионной модели речевого тракта и корректировки её параметров. Метод основан на двухэтапном алгоритме трансформации авторегрессионной модели речевого тракта. На первом этапе тестируют устойчивость авторегрессионной модели по импульсной характеристике формирующего фильтра. На втором этапе при нарушении устойчивости авторегрессионной модели модифицируют импульсную характеристику путем её поэлементного умножения на переменную экспоненциальную величину, которая асимптотически сходится к нулю. Разработан регулярный алгоритм перерасчёта модифицированной импульсной характеристики в откорректированный вектор авторегрессионных параметров на втором этапе трансформации. По результатам экспериментальной апробации предложенного метода сделан вывод о достижении гарантированной устойчивости авторегрессионной модели речевого тракта при её минимальных искажениях в частотной области. Полученные результаты полезны при разработке и модернизации систем автоматического распознавания речи, цифровой речевой связи, искусственного интеллекта и других информационных систем, использующих сжатие данных и кодирование речи на основе авторегрессионной модели речевого тракта при автоматической обработке речевого сигнала.

Within the framework of the traditional direction of research in the field of acoustic measurements, an autoregressive model of the vocal tract as a key link in the human speech apparatus is considered. The acute problem of ensuring the stability of the autoregressive model in systems with adaptation of its parameters to the observed speech signal of short duration is pointed out. To overcome this problem, the task was set of testing the stability of the autoregressive model and adjusting its parameters based on the results of this testing. The study is based on the author’s method of formant analysis of vowel sounds of speech through the synthesis of a recursive shaping filter in the free oscillation mode. To solve sated task, a method is proposed for testing the stability and adjusting the parameters of the autoregressive model of the vocal tract based on a two-stage algorithm for its transformation. At the first stage of transformation, the stability of the autoregressive model is tested using the impulse response of the shaping filter. At the second stage, if the stability of the autoregressive model is violated, its impulse response is modified by element-by-element multiplication by a variable exponential value that asymptotically converges to zero. A regular algorithm has been developed for recalculating the modified impulse response into an adjusted vector of autoregressive parameters at the second stage of transformation. Based on the results of experimental testing of the proposed method, it was concluded that guaranteed stability of the autoregressive model of the vocal tract has been achieved with minimal distortion in the frequency domain. The results obtained are useful in the development and modernization of automatic speech recognition systems, digital speech communications, artificial intelligence and other information systems that use data compression and speech coding based on an autoregressive model of the vocal tract in automatic speech signal processing.

речевой сигналречевой трактдискретное спектральное моделированиеавторегрессионная модельустойчивость авторегрессионной моделитестирование устойчивостикорректировка параметровspeech signalvocal tractdiscrete spectral modelingautoregressive modelstability of autoregressive modelstability testingparameter adjustmentReferencesTernström S. Special Issue on current trends and future directions in voice acoustics measurement. Applied Sciences, 13(6), 3514, (2023). https://doi.org/10.3390/app13063514Ternström S. Special Issue on Current Trends and Future Directions in Voice Acoustics Measurement. Applied Sciences, 13(6), 3514, (2023). https://doi.org/10.3390/app13063514O’Shaughnessy D. Review of methods for coding of speech signals. Journal on Audio, Speech, and Music Processing, (8), (2023). https://doi.org/10.1186/s13636-023-00274-xO’Shaughnessy D. Review of methods for coding of speech signals. Journal on Audio, Speech, and Music Processing, (8), (2023). https://doi.org/10.1186/s13636-023-00274-xСавченко В. В. Мера различий речевых сигналов по тембру голоса. Измерительная техника, (10), 63–69 (2023). https://doi.org/10.32446/0368-1025it.2023-10-63-69Savchenko V. V. A measure of differences in speech signals by the voice timbre. Measurement Techniques, 66(3), 803–812 (2024). https://doi.org/10.1007/s11018-024-02294-1Rabiner L. R., Shafer R. W. Theory and Applications of Digital Speech Processing. Pearson, Boston (2010).Rabiner L. R., Shafer R. W. Theory and Applications of Digital Speech Processing, Pearson, Boston (2010).Gibson J. Mutual Information, the Linear Prediction Model and CELP Voice Codecs. Information, 10(5), 179 (2019). https://doi.org/10.3390/info10050179Gibson J. Mutual Information, the Linear Prediction Model and CELP Voice Codecs. Information, 10(5), 179 (2019). https://doi.org/10.3390/info10050179Савченко В. В., Савченко Л. В. Метод асинхронного анализа голосового источника речи на основе двухуровневой авторегрессионной модели речевого сигнала. Измерительная техника, (2), 55–62 (2024). https://doi.org/10.32446/0368-1025it. 2024-2-55-62Savchenko V. V., Savchenko L. V. Method for asynchronous analysis of a glottal source based on a two-level autoregressive model of the speech signal. Izmeritel’naya Tekhnika, 73(2), 55–62 (2024). https://doi.org/10.32446/0368-1025it. 2024-2-55-62 (In Russ.)Kim H. S. Linear predictive coding is all-pole resonance modeling, Center for Computer Research in Music and Acoustics, Stanford University (2023).Kim H. S. Linear predictive coding is all-pole resonance modeling, Center for Computer Research in Music and Acoustics, Stanford University (2023).Савченко В. В. Метод сравнительного тестирования параметрических оценок спектра мощности: спектральный анализ через синтез временно́ го ряда. Измерительная техника, (6), 56–62 (2023). https://doi.org/10.32446/0368-1025it.2023-6-56-62Savchenko V. V. Method for comparison testing of parametric power spectrum estimates: spectral analysis via time series synthesis. Measurement Techniques, 66(6), 430–438 (2023). https://doi.org/10.1007/s11018-023-02244-3Savchenko V. V. Method for reduction of speech signal autoregression model for speech transmission systems on lowspeed communication channels. Radioelectronics and Communications Systems, 64, 592–603 (2021). https://doi.org/10.3103/S0735272721110030Savchenko V. V. Method for reduction of speech signal autoregression model for speech transmission systems on lowspeed communication channels. Radioelectronics and Communications Systems, 64, 592–603 (2021). https://doi.org/10.3103/S0735272721110030Kathiresan Th., Maurer D., Suter H., Dellwo V. Formant pattern and spectral shape ambiguity in vowel synthesis: The role of fundamental frequency and formant amplitude. The Journal of Acoustical Society of America, 143(3), 1919–1920 (2018). https://doi.org/10.1121/1.5036258Kathiresan Th., Maurer D., Suter H., Dellwo V. Formant pattern and spectral shape ambiguity in vowel synthesis: The role of fundamental frequency and formant amplitude. The Journal of Acoustical Society of America, 143(3), 1919–1920 (2018). https://doi.org/10.1121/1.5036258Palaparthi A., Titze I. R. Analysis of glottal inverse filtering in the presence of source-filter interaction. Speech Communication, 123, 98–108 (2020). https://doi.org/10.1016/j.specom.2020.07.003Palaparthi A., Titze I. R. Analysis of glottal inverse filtering in the presence of source-filter interaction. Speech Communication, 123, 98–108 (2020). https://doi.org/10.1016/j.specom.2020.07.003Venkatraman A. Algorithms and Software for Predictive and Perceptual Modeling of Speech. Springer Cham. (2011). https://doi.org/10.1007/978-3-031-01516-8Venkatraman A. Algorithms and Software for Predictive and Perceptual Modeling of Speech. Springer Cham. (2011). https://doi.org/10.1007/978-3-031-01516-8Alku P., Kadiri S. R., Gowda D. Refining a deep learning-based formant tracker using linear prediction methods. Computer Speech & Language, 81, 101515 (2023). https://doi.org/10.1016/j.csl.2023.101515Alku P., Kadiri S. R., Gowda D. Refining a deep learning-based formant tracker using linear prediction methods. Computer Speech & Language, 81, 101515 (2023). https://doi.org/10.1016/j.csl.2023.101515Fu M., Wang X., Wang J. Polynomial-Decomposition-Based LPC for Formant Estimation. IEEE Signal Processing Letters, 29, 1392–1396 (2022). https://doi.org/10.1109/LSP.2022.3181523Fu M., Wang X., Wang J. Polynomial-Decomposition-Based LPC for Formant Estimation. IEEE Signal Processing Letters, 29, 1392–1396 (2022). https://doi.org/10.1109/LSP.2022.3181523Candan С. Making linear prediction perform like maximum likelihood in gaussian autoregressive model parameter estimation. Signal Processing, 166, 107256 (2020). https://doi.org/10.1016/j.sigpro.2019.107256Candan С. Making Linear Prediction Perform Like Maximum Likelihood in Gaussian Autoregressive Model Parameter Estimation. Signal Processing, 166, 107256 (2020). https://doi.org/10.1016/j.sigpro.2019.107256Wei B., Gibson J. D. A new discrete spectral modeling method and an application to CELP coding, In: IEEE Signal Processing Letters, 10(4), 101–103 (2003). https://doi.org/10.1109/LSP.2003.808550Wei B., Gibson J. D. A new discrete spectral modeling method and an application to CELP coding, In: IEEE Signal Processing Letters, 10(4), 101–103 (2003). https://doi.org/10.1109/LSP.2003.808550Sadhu S., Hermansky H. Radically Old Way of Computing Spectra: Applications in End-to-End ASR. Audio and Speech Processing (eess.AS); Sound (cs.SD). arXiv:2103.14129 [eess.AS] (2021). https://doi.org/10.48550/arXiv.2103.14129Sadhu S., Hermansky H. Radically Old Way of Computing Spectra: Applications in End-to-End ASR. Audio and Speech Processing (eess.AS); Sound (cs.SD). arXiv:2103.14129 [eess.AS] (2021). https://doi.org/10.48550/arXiv.2103.14129Oh H. Recursively Adaptive Randomized Multi-Tree Coding (RAR MTC) of Speech with VAD/CNG. University of California, Santa Barbara, Theses and Dissertations (2023).Oh H. Recursively Adaptive Randomized Multi-Tree Coding (RAR MTC) of Speech with VAD/CNG. University of California, Santa Barbara, Theses and Dissertations (2023).Marple S. L. Digital Spectral Analysis with Applications. 2nd ed., Dover Publications, Mineola, New York (2019).Marple S. L. Digital Spectral Analysis with Applications. 2nd ed., Dover Publications, Mineola, New York (2019).Cui S., Li E., Kang X. Autoregressive Model Based Smoothing Forensics of Very Short Speech Clips. 2020 IEEE International Conference on Multimedia and Expo (ICME), London, UK, рр. 1–6 (2020). https://doi.org/10.1109/ICME46284.2020.9102765Cui S., Li E., Kang X. Autoregressive model based smoothing forensics of very short speech clips. In: 2020 IEEE International Conference on Multimedia and Expo (ICME), London, UK, 1–6 (2020). https://doi.org/10.1109/ICME46284.2020.9102765Esfandiari M., Vorobyov S. A., Karimi M., New estimation methods for autoregressive process in the presence of white observation noise. Signal Processing, 171, 107480 (2020). https://doi.org/10.1016/j.sigpro.2020.107480Esfandiari M., Vorobyov S. A., Karimi M., New estimation methods for autoregressive process in the presence of white observation noise. Signal Processing, 171, 107480 (2020). https://doi.org/10.1016/j.sigpro.2020.107480Савченко В. В., Савченко Л. В. Субоптимальный алгоритм измерения частоты основного тона с использованием дискретного фурье-преобразования речевого сигнала. Радиотехника и электроника, 68(7), 660–668 (2023). https://doi.org/10.31857/S0033849423060128Savchenko V. V., Savchenko L. V. Suboptimal algorithm for measuring pitch frequency using discrete fourier transform of a speech signal. Journal of Communications Technology and Electronics, 68(7), 757–764 (2023). https://doi.org/10.1134/S1064226923060128O’Shaughnessy D. Review of analysis methods for speech applications. Speech Communication, 151, 64–75 (2023). https://doi.org/10.1016/j.specom.2023.05.008O’Shaughnessy D. Review of analysis methods for speech applications. Speech Communication, 151, 64–75 (2023). https://doi.org/10.1016/j.specom.2023.05.008Ternström S., Pabon P. Voice Maps as a Tool for Understanding and Dealing with Variability in the Voice. Applied Sciences, 12, 11353 (2022). https://doi.org/10.3390/app122211353Ternström S., Pabon P. Voice Maps as a Tool for Understanding and Dealing with Variability in the Voice. Applied Sciences, 12, 11353 (2022). https://doi.org/10.3390/app122211353Sun P., Mahdi A., Xu J., Qin J. Speech enhancement in spectral envelope and details subspaces. Speech Communication, 101, 57–69 (2018). https://doi.org/10.1016/j.specom.2018.05.006Sun P., Mahdi A., Xu J., Qin J. Speech enhancement in spectral envelope and details subspaces. Speech Communication, 101, 57–69 (2018). https://doi.org/10.1016/j.specom.2018.05.006Tohyama M. Spectral envelope and source signature analysis. In: Acoustic signals and hearing. Academic Press, 89– 110 (2020). https://doi.org/10.1016/B978-0-12-816391-7.00013-9Tohyama M. Spectral envelope and source signature analysis. In: Acoustic signals and hearing. Academic Press, 89– 110 (2020). https://doi.org/10.1016/B978-0-12-816391-7.00013-9Савченко В. В. Метод авторегрессионного моделирования речевого сигнала с использованием огибающей периодограммы Шустера в качестве опорного спектрального образца. Радиотехника и электроника, 68(2), 138–145 (2023). https://doi.org/10.31857/S0033849423020122Savchenko V. V. A Method for autoregression modeling of a speech signal using the envelope of the schuster periodogram as a reference spectral sample. Journal of Communications Technology and Electronics, 68(2), 121–127 (2023). https://doi.org/10.1134/S1064226923020122El-Jaroudi A., Makhoul J. Discrete all-pole modeling. IEEE Transactions on Signal Processing, 39(2), 411–423 (1991). https://doi.org/10.1109/78.80824El-Jaroudi A., Makhoul J. Discrete all-pole modeling. IEEE Transactions on Signal Processing, 39(2), 411–423 (1991). https://doi.org/10.1109/78.80824Mustiere F., Bouchard M., Bolic M. All-Pole modeling of discrete spectral powers: a unified approach. IEEE Transactions on Audio Speech and Language Processing, 20(2), 705–708 (2012). https://doi.org/10.1109/TASL.2011.2163511Mustiere F., Bouchard M., Bolic M. All-Pole modeling of discrete spectral powers: a unified approach. IEEE Transactions on Audio Speech and Language Processing, 20(2), 705–708 (2012). https://doi.org/10.1109/TASL.2011.2163511Савченко В. В., Савченко Л. В. Метод авторегрессионного моделирования речевого сигнала на основе его дискретного Фурье-преобразования и масштабно-инвариантной меры информационного рассогласования. Радиотехника и электроника, 66(11), 1100–1108 (2021). https://doi.org/10.31857/S0033849421110085Savchenko V. V., Savchenko L. V. Speech signal autoregression modeling based on the discrete fourier transform and scale-invariant measure of information discrimination. Journal of Communications Technology and Electronics, 66(11), 1266–1273 (2021). https://doi.org/10.1134/s1064226921110085Vinay H., Lavanya P., Hippargi A. A., Purohith A., Lohith D. A comparative analysis on speech enhancement and coding techniques. In: 2021 International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT), Bangalore, India, 543–549 (2021). https://doi.org/10.1109/RTEICT52294.2021.9573847Vinay H., Lavanya P., Hippargi A. A., Purohith A., Lohith D. A comparative analysis on speech enhancement and coding techniques. In: 2021 International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT), Bangalore, India, 543–549 (2021). https://doi.org/10.1109/RTEICT52294.2021.9573847Савченко В. В. Метод фонетического декодирования слов с подавлением фонового шума. Радиотехника и электроника, 62(7), 76–80 (2017). https://doi.org/10.7868/S0033849417070099Savchenko V. V. Words phonetic decoding method with the suppression of background noise. Journal of Communications Technology and Electronics, 62(7), 788–793 (2017). https://doi.org/10.1134/S1064226917070099Palani S. Principles of digital signal processing. 2nd Edition. Springer Cham. (2022). https://doi.org/10.1007/978-3-030-96322-4Palani S. Principles of digital signal processing. 2nd Edition. Springer Cham. (2022). https://doi.org/10.1007/978-3-030-96322-4Nam S. H. Stabilizing discrete spectral modeling of audio signals. IEEE Signal Processing Letters, 9(9), 292–294 (2002). https://doi.org/10.1109/LSP.2002.803406Nam S. H. Stabilizing discrete spectral modeling of audio signals. IEEE Signal Processing Letters, 9(9), 292–294 (2002). https://doi.org/10.1109/LSP.2002.803406Magi C., Pohjalainen J., Backstrom T., Alku P. Stabilised weighted linear prediction. Speech Communication, 51(5), 401–411 (2009). https://doi.org/10.1016/j.specom.2008.12.005Magi C., Pohjalainen J., Backstrom T., Alku P. Stabilised weighted linear prediction. Speech Communication, 51(5), 401–411 (2009). https://doi.org/10.1016/j.specom.2008.12.005Miran K. S., Pal P., Babadi B., Wu M. Sampling requirements for stable autoregressive estimation. IEEE Transactions on Signal Processing, 65(9), 2333–2347 (2017). https://doi.org/10.1109/TSP.2017.2656848Miran K. S., Pal P., Babadi B., Wu M. Sampling requirements for stable autoregressive estimation. IEEE Transactions on Signal Processing, 65(9), 2333–2347 (2017). https://doi.org/10.1109/TSP.2017.2656848Kumar S., Singh S. K., Bhattacharya S. Performance evaluation of a ACF-AMDF based pitch detection scheme in realtime. International Journal of Speech Technology, 18, 521–527 (2015). https://doi.org/10.1007/s10772-015-9296-2Kumar S., Singh S. K., Bhattacharya S. Performance evaluation of a ACF-AMDF based pitch detection scheme in realtime. International Journal of Speech Technology, 18, 521–527 (2015). https://doi.org/10.1007/s10772-015-9296-2

The authors declare that there are no conflicts of interest present.