bsuirДоклады БГУИРDoklady BGUIR1729-76482708-0382БГУИР10.35596/1729-7648-2020-18-2-23-29bsuir-2637Research ArticleЭЛЕКТРОНИКА, РАДИОФИЗИКА, РАДИОТЕХНИКА, ИНФОРМАТИКАELECTRONICS, RADIOPHYSICS, RADIOENGINEERING, INFORMATICSAN EFFICIENT SPEECH GENERATIVE MODEL BASED ON DETERMINISTIC/STOCHASTIC SEPARATION OF SPECTRAL ENVELOPESAN EFFICIENT SPEECH GENERATIVE MODEL BASED ON DETERMINISTIC/STOCHASTIC SEPARATION OF SPECTRAL ENVELOPESTahaM.TahaM.

Minsk

Mostafa Taha, Master of Sciences, PhD student of Computer Engineering Department

Minsk

AzarovE. S.AzarovE. S.

Minsk

Elias S. Azarov, D.Sci., Professor of Computer Engineering Department

Minsk

LikhachovD. S.LikhachovD. S.

Minsk

Likhachov Denis Sergeevich, PhD, Associate Professor of Computer Engineering Department

220013, Republic of Belarus, Minsk, P. Brovki str., 6; tel. +375172938805

likhachov@bsuir.byPetrovskyA. А.PetrovskyA. A.

Minsk

Alexander A. Petrovsky, D.Sci., Professor of Computer Engineering Department

Minsk

Belarusian State University of Informatics and RadioelectronicsBelarusian State University of Informatics and Radioelectronics2020310320201822329Copyright © Taha M., Azarov E.S., Likhachov D.S., Petrovsky A.А., 20202020Taha M., Azarov E.S., Likhachov D.S., Petrovsky A.А.Taha M., Azarov E.S., Likhachov D.S., Petrovsky A.A.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://doklady.bsuir.by/jour/article/view/2637

The paper presents a speech generative model that provides an efficient way of generating speech waveform from its amplitude spectral envelopes. The model is based on hybrid speech representation that includes deterministic (harmonic) and stochastic (noise) components. The main idea behind the approach originates from the fact that speech signal has a determined spectral structure that is statistically bound with deterministic/stochastic energy distribution in the spectrum. The performance of the model is evaluated using an experimental low-bitrate wide-band speech coder. The quality of reconstructed speech is evaluated using objective and subjective methods. Two objective quality characteristics were calculated: Modified Bark Spectral Distortion (MBSD) and Perceptual Evaluation of Speech Quality (PESQ). Narrow-band and wide-band versions of the proposed solution were compared with MELP (Mixed Excitation Linear Prediction) speech coder and AMR (Adaptive Multi-Rate) speech coder, respectively. The speech base of two female and two male speakers were used for testing. The performed tests show that overall performance of the proposed approach is speaker-dependent and it is better for male voices. Supposedly, this difference indicates the influence of pitch highness on separation accuracy. In that way, using the proposed approach in experimental speech compression system provides decent MBSD values and comparable PESQ values with AMR speech coder at 6,6 kbit/s. Additional subjective listening testsdemonstrate that the implemented coding system retains phonetic content and speaker’s identity. It proves consistency of the proposed approach.

The paper presents a speech generative model that provides an efficient way of generating speech waveform from its amplitude spectral envelopes. The model is based on hybrid speech representation that includes deterministic (harmonic) and stochastic (noise) components. The main idea behind the approach originates from the fact that speech signal has a determined spectral structure that is statistically bound with deterministic/stochastic energy distribution in the spectrum. The performance of the model is evaluated using an experimental low-bitrate wide-band speech coder. The quality of reconstructed speech is evaluated using objective and subjective methods. Two objective quality characteristics were calculated: Modified Bark Spectral Distortion (MBSD) and Perceptual Evaluation of Speech Quality (PESQ). Narrow-band and wide-band versions of the proposed solution were compared with MELP (Mixed Excitation Linear Prediction) speech coder and AMR (Adaptive Multi-Rate) speech coder, respectively. The speech base of two female and two male speakers were used for testing. The performed tests show that overall performance of the proposed approach is speaker-dependent and it is better for male voices. Supposedly, this difference indicates the influence of pitch highness on separation accuracy. In that way, using the proposed approach in experimental speech compression system provides decent MBSD values and comparable PESQ values with AMR speech coder at 6,6 kbit/s. Additional subjective listening testsdemonstrate that the implemented coding system retains phonetic content and speaker’s identity. It proves consistency of the proposed approach.

speech generative modelharmonic plus noise modelspeech analysisspeech codingspeech generative modelharmonic plus noise modelspeech analysisspeech codingReferencesA. van den Oord, Dieleman S., Zen H., Simonyan K., Vinyals O., Graves A., Kalchbrenner N., Senior A., Kavukcuoglu K. WaveNet: A generative model for raw audio, arXiv:1609.03499, 2016.A. van den Oord, Dieleman S., Zen H., Simonyan K., Vinyals O., Graves A., Kalchbrenner N., Senior A., Kavukcuoglu K. WaveNet: A generative model for raw audio, arXiv:1609.03499, 2016.Shen J., Pang R., Weiss R. J., Schuster M., Jaitly N., Yang Z., Chen Z., Zhang Y., Wang Y., Skerrv-Ryan R. “Natural TTS synthesis by conditioning WaveNet on mel spectrogram predictions,” in Proc. International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2018:4779-4783.Shen J., Pang R., Weiss R. J., Schuster M., Jaitly N., Yang Z., Chen Z., Zhang Y., Wang Y., Skerrv-Ryan R. “Natural TTS synthesis by conditioning WaveNet on mel spectrogram predictions,” in Proc. International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2018:4779-4783.Arik S., Diamos G., Gibiansky A., Miller J., Peng K., Ping W., Raiman J., and Zhou Y. Deep voice 2: Multi-speaker neural text-to-speech. arXiv:1705.08947, 2017.Arik S., Diamos G., Gibiansky A., Miller J., Peng K., Ping W., Raiman J., and Zhou Y. Deep voice 2: Multi-speaker neural text-to-speech. arXiv:1705.08947, 2017.Valin J.-V., Skoglund J. LPCNet: Improving neural speech synthesis through linear prediction, arXiv:1810.11846Valin J.-V., Skoglund J. LPCNet: Improving neural speech synthesis through linear prediction, arXiv:1810.11846Griffin D., Lim J. A new model-based speech analysis/synthesis system. In Proc. International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 1985;10:513-516.Griffin D., Lim J. A new model-based speech analysis/synthesis system. In Proc. International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 1985;10:513-516.Laroche J., Stylianou Y., Moulines E. HNS: Speech modification based on a harmonic+noise model. Proceedings of the ICASSP-93 IEEE International Conference on Acoustics, Speech, and Signal Processing, 1993;2:550-553.Laroche J., Stylianou Y., Moulines E. HNS: Speech modification based on a harmonic+noise model. Proceedings of the ICASSP-93 IEEE International Conference on Acoustics, Speech, and Signal Processing, 1993;2:550-553.Serra X. Musical sound modeling with sinusoids plus noise. Musical Signal Processing (C. Roads, S. Popea, A. Picialli, G. De Poli Eds.). Swets & Zeitlinger Publishers; 1997.Serra X. Musical sound modeling with sinusoids plus noise. Musical Signal Processing (C. Roads, S. Popea, A. Picialli, G. De Poli Eds.). Swets & Zeitlinger Publishers; 1997.Azarov E., Petrovsky A. Instantaneous harmonic analysis for vocal processing. Proceedings of DAFx-09. Como, Italy, September 14; 2009.Azarov E., Petrovsky A. Instantaneous harmonic analysis for vocal processing. Proceedings of DAFx-09. Como, Italy, September 14; 2009.

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