A highly scalable FPGA implementation for cross-correlation with up-sampling support

Abstract

Cross-correlation is a fundamental algorithm which is used in many fields of digital signal processing. One important application is sound source localization. For this purpose, actual research shows a significant improvement of localization accuracy when using up-sampling-based algorithms. These state-of-the-art approaches are challenging for embedded realizations as up-sampling in combination with high sample rates and measurement window lengths leads to huge amounts of data to be processed and stored. This chapter presents hardware acceleration techniques for cross-correlation with up-sampling support. The work introduces a novel processing architecture which addresses large-scale memory consumption compared to conventional solutions. The novel design introduces a highly scalable field programmable gate array realization with linear memory complexity and an interpolation factor-independent runtime.