You are here
Multiple speaker localization
Sound source localization aims at deriving an explicit or implicit estimation of the coordinates of an active source, based on the signals acquired by a set of microphones distributed in space. The localization of the speaker in a room is characterized by the wideband nature of the speech signal, the effect of the enclosure, and the interference of other competitive sources. The most common approaches are based on TDOA (time difference of arrival) estimation and acoustic map estimation. The task is particularly complex in highly reverberant and noisy environments, in the presence of one or more interfering speakers and sound sources, and when the targeted sound source is moving.
Related keywords: time-delay estimation, acoustic maps, head orientation, multiple-speaker tracking, particle filtering, impulse response, far-field and near-field, room geometry inference, direct-to-reverberation ratio