Accelerating FCM Algorithm Using High-Speed FPGA Reconfigurable Computing Architecture

Fuzzy C-Means (FCM) algorithm is a clustering algorithm that is frequently used to enhance the detection accuracy for different applications. However, FCM is a high computationally extensive algorithm where different optimization techniques could be utilized to enhance the computation time. Therefore, in this study, the FCM algorithm was implemented and parallelized on field-programmable gate array (FPGA) platform using the Vivado HLS tool to improve the performance in terms of the execution time. Different optimization techniques were adopted and applied such as loop unrolling, loop pipelining, and dataflow optimization techniques which significantly improved the execution time. Further, the experimental results showed that the speedup of the proposed method over the sequential one is 76 times. More speedup is obtained with increasing the number of iterations due to the exploitation of the parallel FPGA platform and constructed the proposed hardware architecture using different optimization techniques.