Time-Frequency Phase Retrieval for Audio—The Effect of Transform Parameters

In audio processing applications, phase retrieval (PR) is often performed from the magnitude of short-time Fourier transform (STFT) coefficients. Although PR performance has been observed to depend on considered STFT parameters and data, extent this dependence not systematically evaluated yet. To address this, we studied three algorithms for various types content such as redundancy, time-frequency ratio, type window. The quality was in terms objective difference grade signal-to-noise ratio magnitude, provide auditory- signal-based assessments. Our results show that improved with increasing a strong relevance ratio. effect smaller but still observable. window only significant one algorithms. Interestingly, good quality, each required different set parameters, demonstrating individual parameter sets fair comparison across Based these results, developed guidelines optimizing given application.