Accelerating Spectrum Sensing on GPGPU
Lead Presenter: Leiming Yu
Faculty Advisor/Principal Investigator: Miriam Leeser
Method of Presentation: Poster
Spectrum Sensing is critical for Software Defined Radio (SDR) in order to capture the utilization status of wireless bands and to minimize interference between channel users. Spectrum Sensing (SS) is followed by transmission of data in an empty wireless band. The goal is to make spectrum sensing fast and permit more time for transmitting. One technique for SS is energy detection using FFT. A threshold is applied to determine if a resulting energy band is occupied. In this research, we use K-means clustering to dynamically set the threshold based on the stregth of the signals in each energy band. To accelerate SS, we implement FFT and clustering on a General Purpose Graphic Processing Units (GPGPU). We evaluate the performance benefit by comparing an NVIDIA GPU implementation with Matlab.