Mono-modal biometric systems suffer from the drawback that their “error rate” on personal recognition could not meet the stringent performance requirements in many real scenarios.
With regard to this problem, the multi-modal “fusion” of biometric systems allows satisfying these requirements more easily. At present, such systems can be categorized into:
- Mono-biometric systems with multiple recognition algorithms;
- Multi-biometric systems;
- Hybrid biometric systems.
PRA Lab concurred to the state-of-the-art advancement on multi-biometric systems based on fingerprint and face biometric traits. Novel algorithms and frameworks for combining information coming from biometric and also soft-biometrics (e.g. hair colour) have been proposed. A tool exploiting the joint use of face and fingerprint biometrics has been recently developed.
Video-surveillance applications have been also taken into account, by the development of systems fusing biometrics and remote localization devices, based on RFID. Since these systems are also aimed to check the behavior of subjects into a certain scene, head-pose estimation algorithms have been also designed.
Multi-algorithmic method based on the novel perceptron-based class-separation loss function proposed by PRA Lab. This system has been tested on fingerprint and face recognition applications.