Module for processing an audio-video stream associating the spoken words with the corresponding faces

摘要

The invention relates to a module for processing an audio-video stream comprising: an audio segmentation sub-module (31) capable of identifying human words in the audio data of said audio-video stream and of dissociating those spoken simultaneously by different people; - a face detection submodule (32) capable of identifying the human faces present in the video data of said audio-video stream; and- an association sub-module (33) capable of associating the words spoken by each person with their face; characterized in that said association sub-module (33) implements an algorithm using a neural network of deep learning previously trained from a learning base. The invention also relates to a personal digital assistant with voice control for a motor vehicle incorporating such a processing module as well as a method for detecting and applying in a contextualized manner voice control requests in a motor vehicle. Figure to be published with the abstract: Fig. 1