基于总错误率和特征关联的自适应融合多模态生物特征识别

针对单生物特征识别准确率和鲁棒性差的问题, 提出了一种基于总错误率(TER)和特征关联自适应融合多模态生物特 征识别方法。首先将TER作为判别特征引入到多模态识别,以代替传统的匹配分 数；其次在不确定度量理论的基 础上,考虑人脸特征和语音特征之间的时空关联性,提出了一种基于特征关联的多特征 自适应融合策略,利用特征关联 系数自适应调节不同识别特征对识别结果的贡献。仿真实验表明,与几种代表性的融合算法 相比,本文所 提出的融合模式可以有效提高多生物特征识别系统的准确性和鲁棒性。

Multibiometric recognition based on total error rate and feature correlation adaptive fusion

On account of the bad accuracy and rob ustness of unimodal biometric recognition,a novel approach is proposed based on total error rate (TER) and feature correlation adaptive fusion method.This study aims at voiding calculating posterior probability of a certain score transfers scores into TER based on corresponding false reject rates (FRRs) and false accept rates (FARs) at first.Secondly,a new adaptive fusion method named based on information e ntropy theory and uncertainty measurement theory is proposed,which considers the correlation between face information and speech information,and the new adaptive fusion method adjusts the fusion weights adapt ively between weight sum fusion sc heme and multiplicative fusion scheme using the characteristic of correlation coefficient.Experimental results show that the proposed method outperforms the traditional score level fusion schemes,the experimental result also confirms the effectiveness of the proposed method to improve the performance of multibiometric system is confirmed.