Exploiting spatio-temporal spectral features of the indoor acoustic field for sound source localization

The estimation and classification of spectral features of the spatio-temporal acoustic field in a reverberant rectangular enclosure is undertaken with application to localizing sound sources. The field is constructed using an image source model of the impulse response that captures the acoustic properties of the enclosure boundaries. The received field is examined considering linear and randomly distributed array configurations and spectral peaks are characterized to derive a set of features for classification and source identification. The effects of noise generated by the environment and the diffuse field from multiple reflections is distinguished.