Authentication is undoubtedly an important task for all systems providing a component responsible for the interaction between humans and computers. Traditional authentication techniques like PINs, passwords or ID cards show significant drawbacks: they might be forgotten, misplaced or lost, or even stolen, copied or forged. Biometrics use physical or behavioral characteristics to verify the identity of a person and thus overcome these problems. In this respect, biometric technology may be easier and more comfortable to use. First, a short overview of biometric technology in general is given. Then, the main part of this chapter explains the biometric technologies integrated in the SmartKom system. In particular, a new approach to combine several biometrics in a multimodal device is presented. The performance of this proposed combination method is shown to be superior to that of the single biometric subsystems.

In this paper we evaluate on a forensic task our text and language independent speaker recognition system, characterized by modest memory requirements and robustness to environment noise. Noise robustness is achieved by employing a Kalman filter-based sequential interacting multiple models (SIMM) algorithm. The evaluation data was provided by the Netherlands Forensic Institute (NFI) and consisted of telephone conversations in four different languages gathered in real police investigations. The results of NFI evaluation show that our small-footprint system provides competitive equal error rates (EER) for the class of text independent systems operating on telephone speech with strong channel mismatch.

Accurate discrimination between speech and non-speech is an essentialpartinmanytasksofspeechprocessingsystems. Inthis paper an approach to the classification part of a Voice Activity Detector (VAD) is presented. Some possible shortcomings of presentVAD-systemsaredescribedandaclassificationapproachwhichovercomestheseweaknessesisderived.ThisapproachisbasedonaSelf-OrganizingMap(SOM),aneuralnetwork,whichisabletodetectclusterswithinthefeaturespaceofits training data. Training of the classifier takes place in two steps: First the SOM has to be trained. When finished, it is used in the second training step to learn the mapping between its classes and the desired output "speech" resp. "non-speech". Experiments on a database containing audio-samples obtained under different noisy conditions show the potential of the proposed algorithm.

This paper presents a new approach to combine several biometrics in a multimodal device. The proposed approach does not use typical fusion strategies but is based on the multi-dimensional cost probability densities of the originals and forgers, which can be measured or calculated from the intrinsic cost or score distribution of the single biometrics, and uses different decision strategies within this distribution. The performance of the proposed combination method is shown to be superior to that of the single biometric subsystems. The main advantage of this method is the selectable degree of security or comfort. This means that the false acceptance rate (FAR) or the false rejection rate (FRR) of the overall system can be set according to the desired requirements of the respective authentication scenario. Furthermore the approach enables the combination and integration of all biometric systems which provide the cost distribution of originals and forgers even if they come from different suppliers.