Blind identification of aggregated microphones in time domain

This paper introduces an algorithm for blind identification of aggregated microphones in time domain. The features of our approach are summarized as follows: (1) The proposed method treats the blind identification problem of anechoic mixtures in the time domain. (2) The proposed method can identify the gain of each microphone for the directions of sounds whose number is more than the number of the microphones. (3) The proposed method does not utilize the statistical independence of the sounds. The sounds may be not only statistically independent but may also be statistically dependent. (4) The proposed method treats the partially disjoint sounds in the time domain. The sounds may overlap in the frequency domain unlike the sparseness approach. (5) The proposed method does not need to estimate the intervals where sounds are disjoint. First, it is shown that the problem of blind identification and blind source separation can be described not as a convolutive model, but as an instantaneous model in the case of the anechoic mixing when aggregated microphones are assumed. The necessary conditions and the algorithm with experimental results are also described.