Improvement of the Spectral Subtraction Method by High-resolution Frequency Analysis

The Nonharmonic Analysis, recently proposed by the authors, is able to obtain voice and noise spectra independent of each other as a result of high frequency resolution. Therefore, NHA may even enable voices in a noisy environment to be extracted from spectra without distortion. For preprocessing by the spectral subtraction method of effective noise suppression, we used NHA instead of the discrete Fourier transform and the effects on noise suppression were quantitatively verified. In an environment of SNR from -10 dB to +10 dB, the proposed technique showed an improvement of about 4 dB on average