Under CopyrightColeman, Graham KeithGraham KeithColeman2025-12-082025-12-082025-12-01https://publica.fraunhofer.de/handle/publica/500332https://doi.org/10.24406/publica-675810.24406/publica-6758This report describes work undertaken to explain the decisions of speaker identification networks, within the framework of the German VIKING Project on Trustworthy AI for Police Applications (BMBF, 2022-2025). It describes the use of feature attributions, or heatmaps over the audio inputs to speaker ID networks, as well as other interactive visualization techniques. The text previously appears as a part of Fraunhofer IDMT’s final report, in German.enexplainable AIinformation visualizationspeaker identificationtrustworthy AI000 Informatik, Informationswissenschaft, allgemeine Werke::000 Informatik, Wissen, Systeme::000 Informatik, Informationswissenschaft, allgemeine Werke400 Sprache::400 Sprachereport