This paper deals with the problem of speaker detection using audio signal processing, which consists in localizing the current position of a talking speaker in a smart room [Maganti 2006]. This task represents the first step of automatic speaker tracking, which is the overall goal of our research work. An estimation of the current position of the talking person is obtained by comparing the left and right signals captured with two cardioids microphones that are placed in the left side and right side of the meeting room, respectively [Neumann 2009]. The distance separating the two microphones is fixed to a constant value L (figure 1). The idea of using two cardioids microphones is inspired from the human ears, which act as two sound sensors, enabling human beings to find the exact direction of a talking person with great precision. The speaker tracking experiments are conducted off-line in a small room without echo elimination. The obtained results show quite good performances by using the proposed approaches and suggest that those techniques could be efficiently utilized in speaker tracking and localisation.

http://scholarpage.org/journal/