Biometric Identification System Based on Voice Recognition Sing Cepstral Coefficients For Spoofing Detection in Telephone Calls

• Autor: Guzman Zumaeta, Alberto Karel
• Formato: artículo
• Fecha de Publicación: 2023
• Institución: Universidad de Lima
• Repositorio: Revistas - Universidad de Lima
• Lenguaje: español
• OAI Identifier: oai:ojs.pkp.sfu.ca:article/6625
• Enlace del recurso: https://revistas.ulima.edu.pe/index.php/Interfases/article/view/6625
• Nivel de acceso: acceso abierto
• Materia: voice biometrics; Mel frequency cepstral coefficients; spoofing prevention; biometría de voz; coeficientes cepstrales en las frecuencias de Mel; prevención de spoofing

Computer crimes in the telematic systems of company’s harm society because they cause a climate of uncertainty in customers, who have the perception that the computer system, in charge of managing the service or product to be consumed, is not so secure as to trust its money or make transactions remotely. One of the most widespread computer crimes is Spoofing, which consists of impersonating the identity of a person or entity. The objective is to implement a voice recognition system as a mobile application to identify cases of voice impersonation by Spoofing through telephone calls. For this purpose, the Mel scale cepstral coefficients (MFCC) were used as a classifier for cleaning anomalies in the audios, as well as back-propagation neural networks for the user identification system that works together within a mobile application. In the tests carried out, the proposed system had a success rate of 83.5% with 20 entities that were designed by the author out of a total of 2000 audios with 100 corresponding audios from each author for the respective research work. It is concluded that the system is successful in the field of security since it has an optimal acceptance rate and must have a robust system for the different types of Spoofing that has been collected in this research work.