A robust multimodal biometric authentication scheme with voice and face recognition

Abstract EN

This paper proposes a multimodal biometric scheme for human authentication based on fusion of voice and face recognition.

For voice recognition, three categories of features (statistical coefficients, cepstral coefficients and voice timbre) are used and compared.

The voice identification modality is carried out using Gaussian Mixture Model (GMM).

For face recognition, three recognition methods (Eigenface, Linear Discriminate Analysis (LDA), and Gabor filter) are used and compared.

The combination of voice and face biometrics systems into a single multimodal biometrics system is performed using features fusion and scores fusion.

This study shows that the best results are obtained using all the features (cepstral coefficients, statistical coefficients and voice timbre features) for voice recognition, LDA face recognition method and scores fusion for the multimodal biometrics system.