Statistical Models in Forensic Voice Comparison

The purpose of forensic voice comparison is to assist a court of law in deciding whether the voices on two (or more) recordings were produced by the same speaker or by different speakers. The known-speaker recording is often an existing recording in which the identity of the speaker is not disputed,...

Descripción completa

Detalles Bibliográficos
Autores: Morrison Geoffrey, Stewart, Enzinger, Ewald, Ramos Castro, Daniel, González Rodríguez, Joaquín, Lozano-Diez, Alicia
Tipo de recurso: capítulo de libro
Fecha de publicación:2020
País:España
Institución:Universidad Autónoma de Madrid
Repositorio:Biblos-e Archivo. Repositorio Institucional de la UAM
Idioma:inglés
OAI Identifier:oai:repositorio.uam.es:10486/741060
Acceso en línea:https://hdl.handle.net/10486/741060
https://dx.doi.org/10.1201/9780367527709
Access Level:acceso abierto
Palabra clave:Forensic voice comparison
Likelihood ratio
Automatic speaker recognition
Signal processing
Statistical modeling
Telecomunicaciones
Descripción
Sumario:The purpose of forensic voice comparison is to assist a court of law in deciding whether the voices on two (or more) recordings were produced by the same speaker or by different speakers. The known-speaker recording is often an existing recording in which the identity of the speaker is not disputed, but sometimes a recording is made specifically for the purpose of conducting a forensic-voice-comparison analysis. Historically and in present practice, several approaches have been used to extract information from voice recordings and different frameworks have been used to draw inferences from that information. A simple automatic voice-activity detector may be based solely on root-mean-square amplitude, and thus simply find louder parts of the recording