Multimodal biometrics for voice and handwriting

Abstract

In this paper a novel fusion approach for combining voice and online signature verification will be introduced. While the matching algorithm for the speaker identification modality is based on a single Gaussian Mixture Model (GMM) algorithm, the signature verification strategy is based on four different distance measurement functions, combined by multialgorithmic fusion. Together with a feature extraction method presented in our earlier work, the Biometric Hash algorithm, they result in four verification experts for the handwriting subsystem. The fusion results of our new subsystem on the multimodal level are elaborated by enhancements to a system, which was previously introduced by us for biometric authentication in HCI scenarios. Tests have been performed on identical data sets for the original and the enhanced system and the first results presented in this paper show that an increase of recognition accuracy can be achieved by our new multialgorithmic approach for the handwriting modality. © IFIP International Federation for Information Processing 2005.