Direct source and early reflections localization using deep deconvolution network under reverbrate environment

This paper proposes a deconvolution-based network (DCNN) model for DOA estimation of direct source and early reflections under reverberate scenarios. Considering that the first-order reflections of the sound source also contain spatial directivity like the direct source, we treat both of them as the sources in the learning process. We use the covariance matrix of high order Ambisonics (HOA) signals in time domain as the input feature of the network, which is concise while contains precise spatial information under reverberate scenarios. Besides, we use the deconvolution-based network for the spatial pseudo-spectrum (SPS) reconstruction in the 2D polar space, based on which the spatial relationship between elevation and azimuth can be depicted. We have carried out a series of experiments based on simulated and measured data under different reverberate scenarios, which prove the robustness and accuracy of the proposed DCNN model.