Audio-Visual Person Recognition Based on Rank Level Fusion and Gaussian Mixture Model

Abstract

Listen

Translate article

In this paper, A new multimodal system for reliable person identification based on rank level fusion and Gaussian Mixture Model is proposed. Our system fuses the information of face and speech under variant conditions. In logistic regression rank level fusion method, its often difficult to compute the fusion weight properly, so a estimate method based on estimating probability density functions under Gaussian Mixture Model is proposed in this paper. This method can achieve a robust estimator of the optimal weighting parameter which can enhance the rank level fusion performance. The experiments on the AMI database indicate the proposed method can enhance the accuracy of the recognition system effectively.