? ?
Notes from a Medium-Sized Island [entries|archive|friends|userinfo]

[ website | My Website ]
[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

[Dec. 19th, 2015|12:04 pm]

Found an idea in the literature I never would have thought of, and it's so simple and brilliant:
take the spectrogram of an audio recording of speech, (being sure to plot frequency on a linear, not logarithmic scale) and take the 2d fourier transform of that, considered as an image. Because you plotted frequency on a linear scale, harmonics of a fundamental frequency show up as oscillations, in the spectrogram, in the y-direction! Low-frequency oscillations in time in the original signal still show up as oscillations in the spectrogram in the x-direction.

From: altamira16
2015-12-20 03:39 pm (UTC)
Cepstral analysis is also a thing.
(Reply) (Thread)
From: eub
2015-12-21 08:36 am (UTC)
Somewhat reminiscent of the techniques for video motion magnification, IIRC.
(Reply) (Thread)