bsuirДоклады БГУИРDoklady BGUIR1729-76482708-0382БГУИР10.35596/1729-7648-2024-22-3-93-100bsuir-3938Research ArticleСтатьиМетод распознавания эмоций в речевом сигнале с использованием машины опорных векторов и надсегментных акустических признаковSpeech Emotion Recognition Method Based on Support Vector Machine and Suprasegmental Acoustic FeaturesКраснопрошинД. В.KrasnoproshinD. V.

магистрант  каф.  электронных  вычислительных средств

220013, г. Минск, ул. П. Бровки, 6

Master’s Student at the Department  of  Electronic Computing  Facilities

220013, Minsk, P. Brovki St., 6

ВашкевичМ. И.VashkevichM. I.

Вашкевич Максим Иосифович, д-р техн. наук, проф. каф. электронных вычислительных средств

220013, г. Минск, ул. П. Бровки, 6

Тел.: +375 17 293-84-78

Vashkevich Maxim Iosifovich, Dr. of Sci. (Tech.), Professor at the Department of Electronic Computing Facilities

220013, Minsk, P. Brovki St., 6

Tel.: +375 17 293-84-78

vashkevich@bsuir.byБелорусский государственный университет информатики и радиоэлектроникиBelarusian State University of Informatics and Radioelectronics20242406202422393100Copyright © Краснопрошин Д.В., Вашкевич М.И., 20242024Краснопрошин Д.В., Вашкевич М.И.Krasnoproshin D.V., Vashkevich M.I.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/3938

Исследована задача распознавания эмоций в речевом сигнале с использованием мел-частотных  кепстральных  коэффициентов  при  помощи  классификатора  на  основе  метода  опорных  векторов. При   проведении экспериментов применялся набор данных RAVDESS. Предложена модель, которая использует 306-компонентный вектор надсегментных признаков в качестве входных данных для классификатора на основе метода опорных векторов. Оценка качества модели проводилась с помощью невзвешенного среднего значения полноты (UAR). Рассмотрено применение в классификаторе на основе метода опорных векторов в качестве ядра линейной, полиномиальной и радиальной базисной функций. Исследовано использование разных размеров фрейма анализа сигнала (от 23 до 341 мс) на этапе извлечения мел-частотных кепстральных коэффициентов. Результаты исследований выявили значительную точность полученной модели (UAR  = 48 %). Предлагаемый подход демонстрирует потенциал для таких приложений, как голосовые помощники, виртуальные агенты и диагностика психического здоровья.

The problem of recognizing emotions in a speech signal using mel-frequency cepstral coefficients using a classifier based on the support vector machine has been studied. The RAVDESS data set was used in the experiments. A model is proposed that uses a 306-component suprasegmental feature vector as input to a support vector machine classifier. Model quality was assessed using unweighted average recall (UAR). The use of linear, polynomial and radial basis functions as a kernel in a classifier based on the support vector machine is considered. The  use of different signal analysis frame sizes (from 23 to 341 ms) at the stage of extracting mel-frequency cepstral coefficients was investigated. The research results revealed significant accuracy of the resulting model (UAR = 48 %). The proposed approach shows potential for applications such as voice assistants, virtual agents, and mental health diagnostics.

голосовой сигналмел-частотные кепстральные коэффициентыизвлечение аудиопризнаковраспознаваниемашинное обучениеvoice signalmel-frequency cepstral coefficientsaudio feature extractionrecognitionmachine learningReferencesIssa D., Demirci F. M., Yazici A. (2020) Speech Emotion Recognition with Deep Convolutional Neural Networks. Biomedical Signal Processing and Control. 59.Issa D., Demirci F. M., Yazici A. (2020) Speech Emotion Recognition with Deep Convolutional Neural Networks. Biomedical Signal Processing and Control. 59.Luna-Jiménez C., Griol D., Callejas Z., Kleinlein R., Montero J. M., Fernández-Martэínez F. (2021) Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning. Sensors. 21 (22), 1–29.Luna-Jiménez C., Griol D., Callejas Z., Kleinlein R., Montero J. M., Fernández-Martэínez F. (2021) Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning. Sensors. 21 (22), 1–29.Xiao H., Li W., Zeng G., Wu Y., Xue J., Zhang J., [et al.] (2022) On-Road Driver Emotion Recognition Using Facial Expression. Appl. Sci. 12 .Xiao H., Li W., Zeng G., Wu Y., Xue J., Zhang J., [et al.] (2022) On-Road Driver Emotion Recognition Using Facial Expression. Appl. Sci. 12 .Sadok S., Leglaive S., Séguier R. (2023) A Vector Quantized Masked Autoencoder for Speech Emotion Recognition. arXiv preprint arXiv. 2304.Sadok S., Leglaive S., Séguier R. (2023) A Vector Quantized Masked Autoencoder for Speech Emotion Recognition. arXiv preprint arXiv. 2304.Bhavan A., Chauhan P., Shah R. R. (2019) Bagged Support Vector Machines for Emotion Recognition from Speech. Knowledge-Based Systems . 184, 1–7.Bhavan A., Chauhan P., Shah R. R. (2019) Bagged Support Vector Machines for Emotion Recognition from Speech. Knowledge-Based Systems . 184, 1–7.Baruah M., Banerjee B. (2022) Speech Emotion Recognition via Generation Using an Attention-Based Variational Recurrent Neural Network. Proc. Interspeech. 4710–4714.Baruah M., Banerjee B. (2022) Speech Emotion Recognition via Generation Using an Attention-Based Variational Recurrent Neural Network. Proc. Interspeech. 4710–4714.Yu C., Tian Q., Cheng F., Zhang S. (2011) Speech Emotion Recognition Using Support Vector Machines. Advanced Research on Computer Science and Information Engineering. Communications in Computer and Information Science. 152.Yu C., Tian Q., Cheng F., Zhang S. (2011) Speech Emotion Recognition Using Support Vector Machines. Advanced Research on Computer Science and Information Engineering. Communications in Computer and Information Science. 152.Huang X., Acero A., Hon H.-W., Foreword By-Reddy R. (2001) Spoken Language Processing: A Guide to Theory, Algorithm, and System Development. Prentice Hall PTR.Huang X., Acero A., Hon H.-W., Foreword By-Reddy R. (2001) Spoken Language Processing: A Guide to Theory, Algorithm, and System Development. Prentice Hall PTR.Hastie T., Tibshirani R., Friedman J. H., Friedman J. H. (2009) The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer.Hastie T., Tibshirani R., Friedman J. H., Friedman J. H. (2009) The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer.On C. K., Pandiyan P. M., Yaacob S., Saudi A. (2006) Mel-Frequency Cepstral Coefficient Analysis in Speech Recognition. In 2006 International Conference on Computing & Informatics. 1–5.On C. K., Pandiyan P. M., Yaacob S., Saudi A. (2006) Mel-Frequency Cepstral Coefficient Analysis in Speech Recognition. In 2006 International Conference on Computing & Informatics. 1–5.Livingstone S. R., Russo F. A. (2018) The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS): A Dynamic, Multimodal Set of Facial and Vocal Expressions in North American English. PloS One. 13 (5).Livingstone S. R., Russo F. A. (2018) The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS): A Dynamic, Multimodal Set of Facial and Vocal Expressions in North American English. PloS One. 13 (5).Luna-Jiménez C., Griol D., Callejas Z., Kleinlein R., Montero J. M., Fernández-Martínez F. (2021) Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning. Sensors. 21.Luna-Jiménez C., Griol D., Callejas Z., Kleinlein R., Montero J. M., Fernández-Martínez F. (2021) Multimodal Emotion Recognition on RAVDESS Dataset Using Transfer Learning. Sensors. 21.

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