基于自适应并行结构的多模态生物特征识别

传统多模态生物特征识别方法当出现生物特征缺失时,识别性能会明显下降。针对此问题,提出一种融合人脸、虹膜和掌纹的自适应并行结构多模态生物识别方法。该方法在设计融合策略时,考虑到所有可能的输入缺失,构造并行结构的融合函数集,在实际应用时根据输入状态自适应的选择融合策略进行识别。实验仿真结果表明该方法既可提高识别可靠性又可实现当有生物特征缺失时的性能稳定。

A multimodal biometric authentication method based on parallel structure

Once a modality is unavailable or missed,the performance of the traditional method will decrease obviously.To overcome the limitation,a multimodal fusion method combining iris,face and palmprint based on adaptive parallel structure is proposed,which are more robust and flexible especially when some of the biometric modalities are unavailable.Experimental results on our multimodal database proved our conclusion.