Verfahren, System und Computerprogramm zur Deep-Fak-Erkennung in Digitalem Videoinhalt

Details

Veröffentlichungs-Nr.

EP4769389

Anmeldetag

27. Dezember 2024

Veröffentlichung

1. Juli 2026

Rechtsraum

EP

Abstract

A method, system and computer program for deepfake detection in video digital content are provided. Comprises obtaining a video digital content of a user; extracting the audio and video thereof; performing a verbal speech-based detection analysis by transcribing the extracted audio into text and checking whether a speech content is coherent, contextually relevant and/or aligns with an expected topic; performing a voice tonality detection analysis by generating a spectrogram of the extracted audio and identifying irregularities or artifacts characteristics of manipulated or synthesized voice in the spectrogram; performing a facial cue detection analysis by analyzing expression or mouth movements of the user by identifying and dynamically tracking a landmark of the user and verifying whether the analyzed expression or mouth movements temporally and contextually align with speech phonetics; and determining that the obtained video digital content comprises deepfake content based on the performed verbal speech-based detection, voice tonality detection, and facial cue detection analysis.

Anmelder

Firma

Telefonica Innovacion Digital SL

Land

🇪🇸 Spanien