The accuracy of direction of arrival estimation tends to degrade under reverberant conditions due to the presence of reflected signal components which are correlated with the direct path. The recently proposed direct-path dominance test provides a means of identifying time-frequency regions in which a single signal path is dominant. By analysing only these regions it was shown that the accuracy of the FS-MUSIC algorithm could be significantly improved. However, for real-time implementation a less computationally demanding localisation algorithm would be preferable. In the present contribution we investigate the direct-path dominance test as a preprocessing step to pseudo-intensity vector-based localisation. A novel formulation of the pseudo-intensity vector is proposed which further exploits the direct path dominance test and leads to improved localisation performance.Index Terms-direction of arrival estimation, spherical harmonic domain, pseudo-intensity vectors
BACKGROUNDDirection-of-Arrival (DOA) estimation (also known as bearingonly source localisation) is a fundamental problem in acoustic signal processing, particularly as a preprocessing step for beamforming and speech dereverberation [1]. In the context of robot audition it is important for estimation to be computationally efficient such that source localisation can be performed in real time and with low latency on relatively low cost hardware. In this paper we assume a spherical microphone array such that processing can be performed in the Spherical Harmonic Domain (SHD). This is advantageous because beampatterns can be created which are independent of the look direction making the system equally adept at localising sources to the rear and above as those to the front.Localisation using Steered Response Power (SRP) measures the output power from a beamformer as it is steered in turn at a grid of possible source directions and selects one or more maxima as DOA(s). Under anechoic conditions the For MUSIC [6] the noise space of the correlation matrix is used to evaluate the spatial spectrum for a grid of possible source directions. The effective rank of the correlation matrix yields the number of independent sources. In principle, exploiting the sparsity of speech, one can perform accurate DOA estimation for multiple sources by computing the MUSIC spectrum independently for each TF-region whose correlation matrix has effective rank one. However, coherent reflections are not independent and so the effective rank is less than the total number of sources and reflections. To improve the identification of TF-regions whose effective rank is one, the coherence test proposed in [7] averages the correlation matrix over time. The Direct-Path Dominance (DPD) test [8] leads to further improvement using frequency smoothing by also averaging the correlation matrix over frequency thus decorrelating coherent reflections.In contrast to SRP and MUSIC, the Pseudo-Intensity Vector (PIV) approach to DOA estimation [9] does not evaluate a cost function over a grid of poss...