The application of nonuniform fast Fourier transform in audio coding

In this paper, we introduce nonuniform fast Fourier transform (NUFFT) into audio coding. NUFFT benefits from the flexible resolution in different frequency range which is unable to be achieved by the common uniform transforms such as FFT, SDFT, and MDCT. So it matches better to the nonlinear property of human auditory system. As a result, the precision of psychoacoustic model and coding efficiency will be improved. To further increase the calculation efficiency, a new inverse transform method is proposed in this paper. Although some calculation errors are inevitable, fortunately the errors exist only in high frequency range, to which human ear is insensitive. Comparing with FFT, it is shown that NUFFT has better stability to the quantization errors especially at low and middle frequency range. Experiments prove the superiority of NUFFT in both the numerical precision and the objective audio quality, especially at low bit rates.