Parallel wavelet transform based wideband Direction-Of-Arrival (DOA) on multicore/GPU

Abstract

In most of the signal processing applications, main interest is to obtain an accurate estimation of the Direction-Of-Arrival (DOA) of a received signal. Computation of Direction-Of-Arrival (DOA) based on Wavelet transform and other techniques are available in the literature. This work is an improved work of our previously parallel implementation of DOA. In this work, Parallel Haar Wavelet Transform (PHWT) is proposed for use in wideband DOA algorithm. This algorithm has been parallelized and implemented on Intel Nehalem Quad Core processor and NVidia GeForce GTX 260 GPU. The algorithm uses data level and instruction level parallelism for faster execution. The parallel algorithm is developed in C# using .Net framework 4.0 for Multi-Core. It uses combination of C and CUDA for its implementation on GPU. This brings the significant performance on a consumer level PC without any extra cost.