"Human perception is also highly sensitive to discontinuities and irregularities in periodic waveforms. Figure 1 shows that when the stride of the frames does not exactly equal a waveform’s periodicity, the alignment (phase) of the two precesses over time. This condition is assured as at any time there are typically many different frequencies in a given signal. This is a challenge for a synthesis network, as it must learn all the appropriate frequency and phase combinations and activate them in just the right combination to produce a coherent waveform. This phase precession is exactly the same phenomena observed with a short-time Fourier transform (STFT), which is composed of strided filterbanks just like convolutional networks. Phase precession also occurs in situations where filterbanks overlap (window or kernel size < stride)."

Can anybody provide any theory to understand this phenomenon found in STFT and convolution?

 

The Short-time Fourier change (STFT), is a Fourier-related change used to decide the sinusoidal recurrence and stage substance of nearby areas of a sign as it changes over time.[1] practically speaking, the method for figuring STFTs is to partition a more drawn out time signal into shorter portions of equivalent length and afterward process the Fourier change independently on each shorter section. This uncovers the Fourier range on each shorter section. One at that point as a rule plots the changing spectra as a component of time, known as a spectrogram or cascade plot.