This framework facilitates short-time speech processing. A vector containing audio data samples is taken as input, the short-time Fourier transform (STFT) is then computed and the resulting complex spectrum is decomposed into magnitude and phase spectra. No modification is made in this 'hello world' example. Magnitude and phase spectrum components are then recombined. The resulting spectrum is processed through the inverse short-time Fourier analysis. The output signal can then be synthesised using one of two overlap-add approaches: first by 'Allen & Rabiner' and second by 'Griffin & Lim'. A test framework is provided to show the functionality. Enjoy!
Inspired: Phase Spectrum Compensation