Automatic video stream selection method by on-air microphone detection

Florent Lefevre (1, 2) , Vincent Bombardier (1) , Nicolas Krommenacker (1) , Patrick Charpentier (1) , Bertrand Petat (2)

1 CRAN - Centre de Recherche en Automatique de Nancy
2 CitizenCam

This article presents an automatic video editing method for video stream selection in a multi-camera environment. The specific context of this study is council meetings recording and broadcasting. In order to offer the best view to spectator our method is based on a speaker detection, to select the right camera. Since no sound information is available, the proposed method is based on the detection of the change in the visual state of the microphones LED in image sequences, in order to automatically and efficiently select the camera where the speaker is. Studies about the suitable size of the used sliding window and about the relevant features' selection for the verification of microphones' activation are also presented. We have selected seven features to effectively train one classifier, which can be used on different cameras. The feasibility of this approach is shown by the experimentation on councils' videos where the proposed method allows a very efficient detection of the speaker in real-time.

Mots clés: Automatic video editing; Speaker Detection; Feature selection; LED detection

Domaines: Traitement du signal et de l'image [eess.SP]; Vision par ordinateur et reconnaissance de formes [cs.CV]

Florent Lefevre, Vincent Bombardier, Nicolas Krommenacker, Patrick Charpentier, Bertrand Petat. Automatic video stream selection method by on-air microphone detection. International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2018, May 2018, Montréal, Canada. ⟨hal-01798445⟩