Biometrische Sprecher-Erkennung unter unkontrollierbaren Einflussfaktoren im multi-modalen Einsatz auf mobilen Endgeräten (BioMobile)

Project sponsor: LOEWE (Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz)

Project partners: Hochschule Darmstadt, atip GmbH, Bundeskriminalamt (BKA)

Time frame: January 2015 – December 2016

Project description: Facing the increasing interest of companies towards the Bring-Your-Own-Device concept, business-sensitive data needs to be stored secure by easy and user-friendly mechanisms. This can be effectively coped by biometrics providing end-users a low-effort authentication without the need of e.g., remembering passwords. Commonly mobile devices are used, which comprise microphones and touch-sensitive, hence speaker and online signature recognition are focused by this project, where special emphasis is put on speaker recognition. Using mobile devices, biometric voice samples are influenced by various influences, such as changing environmental background noises, the spoken content, or the sample duration. Since these aspects are similar to the forensic speaker recognition case, the German Federal Criminal Police Office (BKA) joins this project as an associated partner. The dual use of this project is expected to yield gains on strengthen the robustness of speaker recognition within the mobile environment, where a multi-modal prototype and public available software modules are anticipated.