In this paper, our aim is to derive a phase spectrum representation computed via the short-time Fourier transform. Specifically, we are interested in developing a narrow-band speech representation - employing 20.40 ms analysis windows. Furthermore, this representation should be as physically meaningful as the magnitude spectrum. To achieve these ends, we concentrate on instantaneous frequency (IF) derived from the phase spectrum. In doing so, we introduce the IF deviation spectrum, and show that this spectrum exhibits pitch and formant structure similar to the magnitude spectrum. Lastly we demonstrate the advantages of the proposed IF deviation spectrum over the IF distribution spectrum proposed earlier in the literature.
Bibliographic reference. Stark, Anthony P. / Paliwal, Kuldip K. (2008): "Speech analysis using instantaneous frequency deviation", In INTERSPEECH-2008, 2602-2605.