Abstract: Multi-modal biometrics techniques have shown more accurately due to the presence of multiple physiological or behavioral characteristics. Multimodal biometrics has become one of inevitable trends in the future. In this paper, D-S fusion algorithm using the recognition rate and the error rate of training set, is proposed. Then through analyzing the recognition rate and error rate, it proposes a modified multi-biometric recognition algorithm based on cumulative frequency and D-S fusion method, named CFDS. The modified D-S algorithm is applied to fusing multi-biometric. Experimental results demonstrate that the modified D-S algorithm is efficient and can improve the reliability of the combination results.

Key words: cumulative frequency, modified D-S theory, multi-modal biometrics recognition

摘要： 多生物特征融合考虑了个体的多种生理或行为特征，因而能显著地改善系统的识别性能，成为生物特征识别技术未来发展趋势之一。利用训练样本的识别率和误识率，提出了基于证据理论的多生物特征融合识别方法；对各识别专家的识别率和误识率进行分析，提出了一种基于累积频率和证据理论（Cumulative Frequency based D-S，CFDS）的多生物特征融合方法；通过几个实验证明了改进的D-S算法的有效性，提高了合成结果的可靠性。

关键词: 累积频率, 改进证据理论, 多生物特征识别