El efecto del enojo en los procesos automatizados de identificación forense de personas locutoras basados en espectros del habla a largo plazo

Manuel Ortega-Rodríguez; Hugo Solís-Sánchez; Diana Valverde-Méndez; Ariadna Venegas-Li

doi:10.55753/aev.v37e54.28

Autores

Manuel Ortega-Rodríguez Escuela de Física y Centro de Investigaciones Geofísicas, Universidad de Costa Rica https://orcid.org/0000-0003-3070-5530
Hugo Solís-Sánchez Escuela de Física y Centro de Investigaciones Geofísicas, Universidad de Costa Rica https://orcid.org/0000-0001-8465-3786
Diana Valverde-Méndez Department of Physics, Princeton University
Ariadna Venegas-Li Physics Department, University of California at Davis https://orcid.org/0000-0002-8660-8513

DOI:

https://doi.org/10.55753/aev.v37e54.28

Palavras-chave:

identificação forense de locutor e locutora, espectros a longo prazo, acústica forense, distorções emocionais, raiva

Resumo

A identificação forense de locutores tem considerado tradicionalmente abordagens ao problema baseadas na análise de espectros a longo prazo (várias dezenas de segundos de duração). Essas abordagens demonstraram ser especialmente robustas, no sentido de que continuam funcionando bem mesmo se as gravações forem curtas; além disso, o método não é sensível a mudanças na intensidade sonora da amostra, e mantém um desempenho adequado na presença de ruído e largura de banda limitada. Por todos esses motivos, constitui uma das técnicas preferidas para a identificação forense, juntamente com a análise de formantes, velocidade da fala e determinação da frequência fundamental. No entanto, verificou-se que o estado de raiva produz uma distorção importante no sinal acústico para efeitos da análise de espectros de fala a longo prazo. Mesmo que o nível de raiva seja apenas moderado, há um desvio dos resultados quantitativos da identificação forense de locutores que representa 33% da distância em direção a um locutor totalmente diferente (no espaço de correlação entre amostras). Portanto, conclui-se que é importante ter cautela ao aplicar este método.

Referências

HOLLIEN, Harry. Barriers to Progress in Speaker Identification with Comments on the Trayvon Martin Case. Linguistic Evidence in Security, Law and Intelligence, University Library System, University of Pittsburgh, v. 1, n. 1, p. 76–98, dic. 2013. ISSN 2327-5596. doi: 10.5195/lesli.2013.3. DOI: https://doi.org/10.5195/LESLI.2013.3

HOLLIEN, Harry. An Approach to Speaker Identification. Journal of Forensic Sciences, Wiley, v. 61, n. 2, p. 334–344, feb. 2016. doi: 10.1111/1556-4029.13034, pMID: 27404606. DOI: https://doi.org/10.1111/1556-4029.13034

HOLLIEN, Harry Francis. Forensic Voice Identification. Londres, Inglaterra: Academic Press, 2002. ISBN 0123526213.

WILLIAMS, Carl E.; STEVENS, Kenneth N. Emotions and speech: Some acoustical correlates. The Journal of the Acoustical Society of America, Acoustical Society of America (ASA), v. 52, n. 4B, p. 1238–1250, oct. 1972. doi: DOI: https://doi.org/10.1121/1.1913238

BANSE, Rainer; SCHERER, Klaus R. Acoustic profiles in vocal emotion expression. Journal of Personality and Social Psychology, American Psychological Association (APA), v. 70, n. 3, p. 614–636, 1996. doi: 10.1037/0022- DOI: https://doi.org/10.1037//0022-3514.70.3.614

JOHNSTONE, Tom. The effect of emotion on voice production and speech acoustics. Tesis (PhD) — University of Western Australia & University of Geneva, Perth, Australia, 2001. doi: https://doi.org/10.31237/osf.io/qd6hz. DOI: https://doi.org/10.31237/osf.io/qd6hz

SCHERER, Klaus R. Voice, Stress, and Emotion. In: . Dynamics of Stress: Physiological, Psychological and Social Perspectives. 1. ed. [S.l.]: Springer US, 1986. p. 157–179. ISBN 978-1-4684-5122-1. doi: 10.1007/978-1-4684- 5122-1_9.

n.◦ 54, diciembre 2022

MARTIN, Maryanne. On the induction of mood. Clinical Psychology Review, Elsevier BV, v. 10, n. 6, p. 669–697, ene. 1990. ISSN 1873-7811. doi: 10.1016/0272-7358(90)90075-l. DOI: https://doi.org/10.1016/0272-7358(90)90075-L

HOLLIEN, Harry; MAJEWSKI, Wojciech. Speaker identification by long-term spectra under normal and distorted speech conditions. The Journal of the Acoustical Society of America, Acoustical Society of America (ASA), v. 62, n. 4, p. 975–980, oct. 1977. ISSN 1520-8524. doi: 10.1121/1.381592. DOI: https://doi.org/10.1121/1.381592

KINNUNEN, Tomi; HAUTAMAKI, Ville; FRANTI, Pasi. On the Use of Long-Term Average Spectrum in Automatic Speaker Recognition. In: Proc. International Symposium on Chinese Spoken Language Processing. [s.n.], 2006. p. 559–567. Disponible en: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=d3b4740466aeb1d25831b6329599b615a5bab9b1.

ORTEGA-RODRIGUEZ, Manuel. Informe Final: Articulación de un sistema de identificación de locutor con fines forenses. [S.l.], 2016. Accedida en noviembre de 2021. Disponible en: https://hdl.handle.net/10669/85190.

HARMEGNIES, Bernard. SDDD: A new dissimilarity index for the comparison of speech spectra. Pattern Recognition Letters, Elsevier BV, v. 8, n. 3, p. 153–158, oct. 1988. ISSN 1872-7344. doi: 10.1016/0167-8655(88)90093-1. DOI: https://doi.org/10.1016/0167-8655(88)90093-1

STANTON, Jeffrey M. Galton, Pearson, and the Peas: A Brief History of Linear Regression for Statistics Instructors. Journal of Statistics Education, Informa UK Limited, v. 9, n. 3, ene. 2001. ISSN 1069-1898. doi: 10.1080/10691898.2001.11910537. DOI: https://doi.org/10.1080/10691898.2001.11910537

FULLER, Fred H. Detection of emotional stress by voice analysis final report. Bethesda, Maryland, USA, 1972. Disponible en: https://www.ojp.gov/ncjrs/virtual-library/abstracts/detection-emotional-stress-voice-analysis-final-report.

HARNSBERGER, James D.; HOLLIEN, Harry; MARTIN, Camilo A.; HOLLIEN, Kevin A. Stress and Deception in Speech: Evaluating Layered Voice Analysis. Journal of Forensic Sciences, Wiley, v. 54, n. 3, p. 642–650, mayo 2009. ISSN 1556-4029. doi: 10.1111/j.1556-4029.2009.01026.x. DOI: https://doi.org/10.1111/j.1556-4029.2009.01026.x

PITTAM, Jeffery. The Long-Term Spectral Measurement of Voice Quality as a Social and Personality Marker: A Review. Language and Speech, SAGE Publications, v. 30, n. 1, p. 1–12, ene. 1987. ISSN 1756-6053. doi: 10.1177/002383098703000101. DOI: https://doi.org/10.1177/002383098703000101

RODMAN, Robert D.; POWELL, Michael S. Computer Recognition of Speakers Who Disguise Their Voice. In: The International Conference on Signal Processing Applications and Technology (ICSPAT 2000). [s.n.], 2000. Disponible en: https://api.semanticscholar.org/CorpusID:16980245.

HERTRICH, I.; ZIEGELMAYER, G. Sexual dimorphism in the long term speech spectrum. Human Evolution, Springer Science and Business Media LLC, v. 2, n. 3, p. 255–262, mayo 1987. doi: 10.1007/bf03016110. DOI: https://doi.org/10.1007/BF03016110

LINVILLE, Sue Ellen. Source Characteristics of Aged Voice Assessed from Long-Term Average Spectra. Journal of Voice, Elsevier BV, v. 16, n. 4, p. 472–479, dic. 2002. doi: 10.1016/s0892-1997(02)00122-4. DOI: https://doi.org/10.1016/S0892-1997(02)00122-4

YÜKSEL, Mustafa; GÜNDÜZ, Bülent. Long term average speech spectra of Turkish. Logopedics Phoniatrics Vocology, Informa UK Limited, v. 43, n. 3, p. 101–105, sep. 2017. doi: 10.1080/14015439.2017.1377286. DOI: https://doi.org/10.1080/14015439.2017.1377286

National Institute of Standards and Technology. NIST/SEMATECH e-Handbook of Statistical Methods. [s.n.], 2012. Accedida en octubre de 2021. Disponible en: https://www.itl.nist.gov/div898/handbook/prc/section2/prc222.htm.

Audacity Team. Audacity (versión 2.1.0), editor y grabador de audio. 2015. Disponible en: https://www.audacityteam.org/.

The International Association for Forensic Phonetics and Acoustics. Code of Practice. [S.l.], 2004. Accedida en enero de 2018. Disponible en: https://www.iafpa.net/the-association/code-of-practice/.