A Hybrid Mel-Frequency Cepstrum Coefficients and Gated Recurrent Units Approach for Voiceprint Recognition
Year: 2024
Type: conference-paper
Abstract: A voiceprint system recognizes an individual using their voice biometrics for person recognition purpose. Voiceprint offers notable advantages in convenience and cost since it only requires a microphone, a ubiquitous component in smartphones, making it highly practical for user identity management such as personalized customer support by recognizing the user using just their voice. This paper proposes a voiceprint recognition system to combine Mel-Frequency Cepstrum Coefficients (MFCC) and Gated Recurrent Units (GRU), which is termed MFCC-GRU method. The resulting mixture is extremely useful in voiceprint recognition, as it could accurately capture and evaluate the unique characteristics and temporal dynamics of an individual's voice for identity verification. The proposed method is tested on three voiceprint datasets and achieving the highest accuracy of 98.17% at best.
Institutions: Multimedia University, University of Nottingham Ningbo China
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4408120426