Forensic comparison of voices: a preliminary study on the differences between a natural voice and an artificial voice for judicial investigation

This study presents an approach to the phonetic similarities and differences between a natural voice and an artificial voice, which is why it seeks to: (a) provide an analysis that serves as a precedent for judicial cases of voice cloning by artificial intelligence (AI) and (b) expose the importance...

Descripción completa

Detalles Bibliográficos
Autores: Jimenez Peña, Jhon, Torres Castillo, Fernando Aarón, Cueva Sanchez, Oscar Esaul
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2024
País:Perú
Institución:Poder Judicial del Perú
Repositorio:Revistas - Poder Judicial del Perú
Idioma:español
OAI Identifier:oai:revistas.pj.gob.pe:article/881
Acceso en línea:https://revistas.pj.gob.pe/revista/index.php/ropj/article/view/881
Access Level:acceso abierto
Palabra clave:phonetic analysis
artificial voice
text to speech converter
forensic linguistics
criminalistics
análisis fonético
voz artificial
convertidor de texto en habla
lingüística forense
criminalística
análise fonética
conversor de texto para fala
linguística forense
Descripción
Sumario:This study presents an approach to the phonetic similarities and differences between a natural voice and an artificial voice, which is why it seeks to: (a) provide an analysis that serves as a precedent for judicial cases of voice cloning by artificial intelligence (AI) and (b) expose the importance of linguistics as a source of scientific evidence for the judicial system. Thus, the voice of the Argentine narrator Mariano Closs and his artificial counterpart created in FakeYou (text-to-speech converter) have been analyzed using the combined method that integrates the use of automatic voice analysis programs (Forensia and SIS II) and the phonetic analysis. The automatic programs showed results of high convergence between the natural voice and the artificial voice. However, in the phonetic analysis, differences were observed in the production of certain sounds, in intonation, and there were phonetic processes present in a sample. Thus, despite the similarity of the samples on the biometric level, the artificial voice of the narrator Mariano Closs is still not completely similar to his natural counterpart on the phonetic level.