Assessment of FPGA Implementations of One Sided Jacobi Algorithm for Singular Value Decomposition

摘要

Nowadays many application domains require an embedded electronic system for tactile data processing. Our research aims to implement a real time embedded electronic system based on tensorial kernel approach for tactile data processing. Singular value decomposition represents the more computational expensive algorithm for the tensorial kernel approach. This paper presents an assessment of FPGA implementations of one sided Jacobi algorithm for singular value decomposition. Designs are implemented to handle an arbitrary m×n matrix with fixed point arithmetic. The results figure out an efficient implementation suitable for real time embedded applications.