An Efficient Implementation Approach to FFT Processor for Spectral Analysis

Jinti Hazarika, Mohd Tasleem Khan*, Shaik Rafi Ahamed, Harshal B. Nemade

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)

Abstract

This article presents an efficient hardware implementation approach to a variable-size fast Fourier transform (FFT) processor for spectral analysis. Due to its capability to handle different frame sizes, it can be adapted in situations where operating parameters necessitate adhering to different standard requirements. A serial real-valued processor with a new data-flow graph is considered, as it requires the least number of multipliers. By joint use of stage-specific optimization and multiplierless structure, the overall hardware efficiency of the proposed design is enhanced. Clock gating is employed to enable the variable-size processor operation along with power reduction. A fixed-point (FP) analysis of the proposed design is considered. The proposed novel multiplierless structure is based on shift and accumulation (SA). This also includes the generation (and sharing) of partial products (PPs) based on their symmetries. The proposed design offers low area and low power as compared with the state of the art. It is demonstrated for spectral analysis of electroencephalogram (EEG) signals for machine-learning-based epileptic seizure prediction on a field-programmable gate array (FPGA) platform.

Original languageEnglish
Article number2006613
JournalIEEE Transactions on Instrumentation and Measurement
Volume72
Early online date4 Aug 2023
DOIs
Publication statusPublished - 2023

Keywords

  • Binary Multiplier
  • Fast Fourier transform (FFT)
  • field-programmable gate array (FPGA)
  • low complexity
  • partial product generator (PPG)
  • spectral analysis

ASJC Scopus subject areas

  • Instrumentation
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'An Efficient Implementation Approach to FFT Processor for Spectral Analysis'. Together they form a unique fingerprint.

Cite this