Method to control multiple cameras in a conference room based on audio tracking and head detection data

Abstract

A method comprises, at a video conference system having cameras and microphone arrays each co-located with a corresponding one of the cameras: detecting a face of a participant, and estimating orientations of the face relative to the cameras, based on video captured by the cameras; receiving, from each of the microphone arrays, at least two microphone signals that represent detected audio from the participant; separately correlating the at least two microphone signals from each of the microphone arrays against each other using a correlation function, to produce correlation peaks for the microphone arrays corresponding to the microphone arrays and the cameras; determining a preferred camera among the cameras based on the correlation peaks and the orientations of the face relative to the cameras; and transmitting the video captured by the preferred camera to a network.