Efficient Sound Localization Algorithm Under Strong Interference

Sound source localization using a drone has demon-strated gratifying results, such as in fault detection, rescue mission, surveillance. At present, the existing classical sound localization algorithms often yield low positioning accuracy. Besides, the high time cost of calculation makes these algorithms difficult to be applied in practice. In this paper, we propose a algorithm that utilizes the invariant property of noise subspace(IPNS) with acceleration on graphics processing units(GPU) to improve the resolution performance and computational efficiency. Sound localization algorithm based on IPNS has severe computational overhead and needs to perform eigenvalue decomposition of thou-sands of matrices. This paper implement QR-Givens algorithm on GPU to solve the time constraint of sound source location problem, which greatly reduces the running time and improves the location efficiency.