Output details
11 - Computer Science and Informatics
University of Surrey
A user-specific and selective multimodal biometric fusion strategy by ranking subjects
<12>Although multimodal biometrics is often thought to require more processing time, we show that reduction in the usage of biometric modalities is possible without compromising system performance, which is useful for systems requiring very high throughput (e.g., busy airports, particularly at peak hours). We have developed a user-specific and selective fusion for combining multiple biometric traits. This approach departs significantly from mainstream fusion methods by using the novel B-ratio criterion to rank subjects based on their match score statistics. The proposed method can achieve similar or better performance than conventional fusion at reduced computational cost.