2011文献TDOA Estimation for Multiple Sound Sources in Noisy and Reverberant Environments Using Broadband Independent Component Analysis.pdfVIP

1490 IEEE TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 19, NO. 6, AUGUST 2011 TDOA Estimation for Multiple Sound Sources in Noisy and Reverberant Environments Using Broadband Independent Component Analysis Anthony Lombard, Yuanhang Zheng, Herbert Buchner, Member, IEEE, and Walter Kellermann, Fellow, IEEE Abstract— In this paper, we show that minimization of the statis- filtering of the acoustic scene. Aside from signal enhancement, tical dependence using broadband independent component anal- the application of microphone arrays for the localization of ysis (ICA) can be successfully exploited for acoustic source localiza- sound sources is of particular interest. Accurate localization of tion. As the ICA signal model inherently accounts for the presence of several sources and multiple sound propagation paths, the ICA one or several sources is often a necessary prerequisite for other criterion offers a theoretically more rigorous framework than con- techniques like, e.g., steering a beamformer [1] or pointing ventional techniques based on an idealized single-path and single- a camera in the direction of a sound source of interest. The source signal model. This leads to algorithms which outperform localization can be performed in the near field, where we are other localization methods, especially in the presence of multiple si- interested in the exact position of each source, or in the far multaneously active sound sources and under adverse conditions, notably in reverberant environments. Three methods are investi- field, where we can only estimate directions of arrival (DOAs), gated to extract the time difference of arrival (TDOA) information disregarding the range coordinat