Audio Spectral Analysis Using the Winograd Fourier Transform

Spectral analysis is the basis of many audio measurement systems, e.g.noise meters, TDS systems, etc.Currently this analysis may be done in real-time via special purpose analogue or digital hardware or in non real-time via software.In many cases a real-time (or near real-time) analysis is necessary and so the instrument designer is forced to grapple with the problems and additional cost of special purpose hardware.It would be desirable to be able to perform real-time spectral analysis using standard microprocessor components as these offer both cost and development advantages.The Hinograd Fourier Transform (NFT) is a discrete Fourier transform (DFT) algorithm which minimises the number of multiplications required to perform the transform by using a more complex sequence of calculations.However, the reduction in the number of multiplications makes the algorithm suitable for implementation in real-time on a standard microprocessor.The rest of this paper describes the NFT algorithm and its implementation.