06ad22e8-1511-4897-a1e7-a8976f18032320251110115604612wseas:wseasmdt@crossref.orgMDT DepositWSEAS TRANSACTIONS ON COMPUTER RESEARCH2415-15211991-875510.37394/232018http://wseas.org/wseas/cms.action?id=13372110202511020251310.37394/232018.2025.13https://wseas.com/journals/cr/2025.phpCesariniValerioDepartment of Electronic Engineering, University of Rome Tor Vergata, Via del Politecnico, 1, 00100 Rome, ITALYCostantiniGiovanniDepartment of Electronic Engineering, University of Rome Tor Vergata, Via del Politecnico, 1, 00100 Rome, ITALYIn machine learning-based voice analysis the sex of the subjects is the most important covariate, as vocal differences between sexes lead to significant discrepancies in acoustic features and model generalization. We propose a signal processing pipeline that normalizes audio recordings of female subjects to exhibit characteristics more similar to those of males, or vice versa. We computed ratios of fundamental frequency, formants, breathiness, and pitch variability from a baseline dataset and used them to build a pipeline based on the decomposition of a signal into fundamental frequency contour, spectral envelope and aperiodicity. We tested on a classification task for the detection of intoxicated versus sober subjects and observed a systematic increase in accuracy when using our processing on the dataset. Accuracies increase up to 5.5% across all three considered algorithms: Support Vector Machine, Random Forest, and Neural Network, with the latter model applied on processed data achieving a state-of-the-art accuracy of 81.79%. Summary: This paper presents a frequency-based audio preprocessing method to reduce gender-related acoustic variability in machine learning-based voice analysis, particularly for detecting alcohol intoxication. The authors developed a pipeline to normalize female voice characteristics toward male standards, addressing differences in fundamental frequency, formants, breathiness, and pitch variability. Applied to a dataset of sober and intoxicated speech samples, this preprocessing significantly improved classification accuracy for intoxication detection, achieving up to a 5.5% accuracy increase, with a neural network reaching a state-of-the-art accuracy of 81.79%. The pipeline effectively minimized sex-based acoustic variability without introducing notable perceptual artifacts, thus enhancing generalization and model reliability across genders in voice-based machine learning tasks.111020251110202566066859https://wseas.com/journals/cr/2025/b205118-014(2025).pdf10.37394/232018.2025.13.59https://wseas.com/journals/cr/2025/b205118-014(2025).pdf10.3390/electronics11233935J. L. Bautista, Y. K. Lee, H. S. Shin, «Speech Emotion Recognition Based on Parallel CNNAttention Networks with Multi-Fold Data Augmentation», Electronics, vol. 11, fasc. 23, Art. fasc. 23, Jan 2022, doi: 10.3390/electronics11233935. 10.3390/s23073461G. Costantini, V. Cesarini, E. Brenna, «HighLevel CNN and Machine Learning Methods for Speaker Recognition», Sensors, vol. 23, fasc. 7, Art. fasc. 7, Jan 2023, doi: 10.3390/s23073461. 10.3390/app142311446Cesarini, V.; Costantini, G. Reverb and Noise as Real-World Effects in Speech Recognition Models: A Study and a Proposal of a Feature Set. Appl. Sci. 2024, 14, 11446. https://doi.org/10.3390/app142311446. 10.1016/j.jvoice.2021.01.018M. Alves, G. Silva, B. C. Bispo, M. E. Dajer, P. M. Rodrigues, «Voice Disorders Detection Through Multiband Cepstral Features of Sustained Vowel», J. Voice, vol. 37, fasc. 3, pp. 322–331, May 2023, doi: 10.1016/j.jvoice.2021.01.018. 10.1007/bf00994018C. Cortes, V. Vapnik, «Support-vector networks», Mach. Learn., vol. 20, fasc. 3, pp. 273–297, Sept. 1995, doi: 10.1007/BF00994018. G. Fant, “Acoustic Theory of Speech Production”. Book, Ed. Walter de Gruyter, 1970. 10.1016/j.protcy.2013.12.124J. P. Teixeira, C. Oliveira, C. Lopes, «Vocal Acoustic Analysis – Jitter, Shimmer and HNR Parameters», Procedia Technol., vol. 9, pp. 1112–1122, 2013, doi: 10.1016/j.protcy.2013.12.124. B. P. Bogert, «The quefrency alanysis of time series for echoes ; Cepstrum, pseudo autocovariance, cross-cepstrum and saphe cracking», Time Ser. Anal., pp. 209–243, 1963. 10.1007/978-981-15-9647-6_39J. Kaur e A. Kumar, «Speech Emotion Recognition Using CNN, k-NN, MLP and Random Forest», 2021, pp. 499–509. doi: 10.1007/978-981-15-9647-6_39. 10.3390/app13158562Cesarini, V.; Saggio, G.; Suppa, A.; Asci, F.; Pisani, A.; Calculli, A.; Fayad, R.; HajjHassan, M.; Costantini, G. Voice Disorder Multi-Class Classification for the Distinction of Parkinson’s Disease and Adductor Spasmodic Dysphonia. Appl. Sci. 2023, 13, 8562. https://doi.org/10.3390/app13158562. 10.1109/icassp.2002.1005790S. Umesh, S. V. Bharath Kumar, M. K. Vinay, R. Sharma and R. Sinha, "A simple approach to non-uniform vowel normalization," 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing, Orlando, FL, USA, 2002, pp. I517-I-520, doi: 10.1109/ICASSP.2002.5743768. 10.1044/jshr.3704.769J. Hillenbrand, R. A. Cleveland, e R. L. Erickson, «Acoustic Correlates of Breathy Vocal Quality», J. Speech Lang. Hear. Res., vol. 37, fasc. 4, pp. 769–778, Aug. 1994, doi: 10.1044/jshr.3704.769. 10.1016/0271-5309(94)00011-zC. Henton, «Pitch dynamism in female and male speech», Lang. Commun., vol. 15, fasc. 1, pp. 43–61, Jan 1995, doi: 10.1016/0271- 5309(94)00011-Z. 10.1371/journal.pone.0156870G. Pino Escobar, J. Terry, B. P. Kriengwatana, P. Escudero, «Speech normalization across speaker, sex and accent variation is handled similarly by listeners of different language backgrounds: Australasian International Conference on Speech Science and Technology», Proc. Sixt. Australas. Int. Conf. Speech Sci. Technol. 6-9 Dec. 2016 Parramatta Aust., pp. 161–164, 2016. 10.1002/9781119184096.ch6K. Johnson, M. J. Sjerps, «Speaker Normalization in Speech Perception», in The Handbook of Speech Perception, John Wiley & Sons, Ltd, 2021, pp. 145–176. doi: 10.1002/9781119184096.ch6. J. D. Miller, «Auditory-perceptual interpretation of the vowel», J. Acoust. Soc. Am., vol. 85, fasc. 5, pp. 2114–2134, 1989, doi: 10.1121/1.397862. 10.1017/s0954394509990160A. H. Fabricius, D. Watt, e D. E. Johnson, «A comparison of three speaker-intrinsic vowel formant frequency normalization algorithms for sociophonetics», Lang. Var. Change, vol. 21, fasc. 3, pp. 413–435, Oct. 2009, doi: 10.1017/S0954394509990160. 10.21437/interspeech.2010-117T. Schaaf, F. Metze, «Analysis of gender normalization using MLP and VTLN features», presented at Proc. Interspeech 2010, pp. 306–309. doi: 10.21437/Interspeech.2010- 117. 10.21437/interspeech.2008-414P. Fousek, L. Lamel, J.-L. Gauvain, «Transcribing broadcast data using MLP features», presented at Proc. Interspeech 2008, pp. 1433–1436. doi: 10.21437/Interspeech.2008-414. L.-M. Zhang, Y. Li, Y.-T. Zhang, G. W. Ng, Y.-B. Leau, H. Yan, «A Deep Learning Method Using Gender-Specific Features for Emotion Recognition», Sensors, vol. 23, fasc. 3, Art. fasc. 3, Jan 2023, doi: 10.3390/s23031355. 10.3390/electronics11101594D. Rizhinashvili, A. H. Sham, G. Anbarjafari, «Gender Neutralisation for Unbiased Speech Synthesising», Electronics, vol. 11, fasc. 10, Art. fasc. 10, Jan 2022, doi: 10.3390/electronics11101594. P. Boersma, D. Weenink, «PRAAT, a system for doing phonetics by computer», Glot Int., vol. 5, pp. 341–345, gen. 2001. «Gender Recognition by Voice(original)». Accessed on: March 15, 2025. [Online]. https://www.kaggle.com/datasets/murtadhanaj im/gender-recognition-by-voiceoriginal 10.1007/s10579-011-9139-yF. Schiel, C. Heinrich, S. Barfüßer, T. Gilg, «ALC: Alcohol Language Corpus», in Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), Marrakech, Morocco: European Language Resources Association (ELRA), May 2008. A. de Cheveigne, H. Kawahara, «YIN, a fundamental frequency estimator for speech and musica)», J Acoust Soc Am, vol. 111, fasc. 4, 2002. 10.1109/tassp.1980.1163489A. Gray, D. Wong, «The Burg algorithm for LPC speech analysis/Synthesis», IEEE Trans. Acoust. Speech Signal Process., vol. 28, fasc. 6, pp. 609–615, dic. 1980, doi: 10.1109/TASSP.1980.1163489. 10.1587/transinf.2015edp7457M. Morise, F. Yokomori, K. Ozawa, «WORLD: A Vocoder-Based High-Quality Speech Synthesis System for Real-Time Applications», IEICE Trans. Inf. Syst., vol. E99.D, fasc. 7, pp. 1877–1884, 2016, doi: 10.1587/transinf.2015EDP7457. 10.21437/interspeech.2017-68M. Morise, «Harvest: A High-Performance Fundamental Frequency Estimator from Speech Signals», in Interspeech 2017, ISCA, ago. 2017, pp. 2321–2325. doi: 10.21437/Interspeech.2017-68. 10.1016/j.specom.2014.09.003«CheapTrick, a spectral envelope estimator for high-quality speech synthesis, Speech Communication, Vol. 67, March 2015, pp. 1- 7. https://doi.org/10.1016/j.specom.2014.09.003. 10.1016/j.specom.2016.09.001M. Morise, «D4C, a band-aperiodicity estimator for high-quality speech synthesis», Speech Commun., vol. 84, pp. 57–65, nov. 2016, doi: 10.1016/j.specom.2016.09.001. 10.1145/2729095.2729097F. Eyben, B. Schuller, «openSMILE:): the Munich open-source large-scale multimedia feature extractor», ACM SIGMultimedia Rec., vol. 6, fasc. 4, pp. 4–13, gen. 2015, doi: 10.1145/2729095.2729097. 10.21437/interspeech.2016-129B. Schuller et al., «The INTERSPEECH 2016 computational paralinguistics challenge: 17th Annual Conference of the International Speech Communication Association, INTERSPEECH 2016», Proc. Annu. Conf. Int. Speech Commun. Assoc. INTERSPEECH, vol. 08-12-September-2016, pp. 2001–2005, 2016, doi: 10.21437/Interspeech.2016-129. 10.1109/89.326616H. Hermansky, N. Morgan, «RASTA processing of speech», IEEE Trans. Speech Audio Process., vol. 2, fasc. 4, pp. 578–589, ott. 1994, doi: 10.1109/89.326616. M. A. Hall, «Correlation-based Feature Selection for Machine Learning», PhD Thesis, University of Waikato, New Zealand, 1999. 10.1201/9781315108230-10M. K. and K. Johnson, “Stepwise Selection”, from Feature Engineering and Selection: A Practical Approach for Predictive Models. Accessed on: February 19, 2023, [Online]. https://bookdown.org/max/FES/greedystepwise-selection.html (Accessed Date: October 10, 2024). 10.1007/bf00116251J. R. Quinlan, «Induction of decision trees», Mach. Learn., vol. 1, fasc. 1, pp. 81–106, mar. 1986, doi: 10.1007/BF00116251. D. P. Kingma e J. Ba, «Adam: A Method for Stochastic Optimization», January 30, 2017, arXiv: arXiv:1412.6980. doi: 10.48550/arXiv.1412.6980. 10.1016/j.eswa.2024.125656F. Amato, V. Cesarini, G. Olmo, G. Saggio, G. Costantini, «Beyond breathalyzers: AIpowered speech analysis for alcohol intoxication detection», Expert Syst. Appl., vol. 262, p. 125656, March 2025, doi: 10.1016/j.eswa.2024.125656. 10.21437/interspeech.2011-805D. Bone, M. Black, M. Li, A. Metallinou, S. Lee, e S. Narayanan, “Intoxicated Speech Detection by Fusion of Speaker Normalized Hierarchical Features and GMM Supervectors.”, Interspeech 2011, p. 3220. doi: 10.21437/Interspeech.2011-805. 10.3389/fpsyg.2024.1412372VL. Holmes, G. Rieger, e S. Paulmann, «The effect of sexual orientation on voice acoustic properties», Front. Psychol., vol. 15, ago. 2024, doi: 10.3389/fpsyg.2024.1412372.