Abstract
Sound Source Localization (SSL) via finding the direction of arrival (DOA) of source signals improves the performance of hearing aid devices (HAD) for the users. In this paper, we present a proposed SSL algorithm - 'Dictionary-based Singular Value Decomposition (SVD) for Principal SL' using the non-uniform non-linear microphone array (NUNLA) available on smartphone as an assistive device to HADs. The proposed algorithm with NUNLA has superior performance compared with uniform linear arrays (ULA) and popular DOA estimation algorithm. Advantages of using proposed SSL method are presented using polar directivity patterns (PDP) and Root Mean Squares Error (RMSE) values using single speech sources in Additive White Gaussian Noise (AWGN) at low signal to noise ratios (SNRs). PDP and RMSE values show that the proposed method, unlike other methods, is capable of accurate SSL over a complete 360° scan, with low average RMSE at very low SNRs.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call