This paper proposes a novel direct source and first-order reflections localization method by integrating the high order Ambisonics (HOA) algorithm and deep neural network. We use the covariance matrix of HOA signals in the time domain as the input feature of the network, which contains precise spatial information of the sound sources under reverberant scenarios. Besides, we use the deconvolution-based neural network (DCNN) for the spatial pseudo-spectrum (SPS) reconstruction, based on which the spatial relationship between elevation and azimuth can be depicted. 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 have carried out a series of experiments based on simulated and measured data under different reverberant scenarios, which prove the effectiveness and accuracy of the proposed DCNN model.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
