Paleospectrophony

NOTE: This initial account of methods for transducing images into sound—first posted in 2009—is maintained here for its historical interest. You can also find it archived in Archive.org's Wayback Machine. For current information, please visit griffonage.com.

1. Introduction

1.1—Paleospectrophony is the name I've given to a distinctive application for reverse Fourier analysis software such as Coagula, AudioPaint and MetaSynth's Image Synth. Each of these programs can "play" any digital image as though it were a spectrogram: one axis is time, the other is frequency, and the intensity of individual pixels corresponds to amplitude. 

Paleospectrophony is the use of this type of software to "play" old inscriptions of sound.

1.2—People have long been using reverse Fourier analysis software as a tool in experimental music, but as far as I can tell, said efforts have all taken one of two forms: (1) playing existing images—e.g. the Mona Lisa, a photograph of a tree—that were not originally created as graphs of time versus frequency or (2) playing spectrographic images created expressly for this purpose.

CEDAR Retouch and the "remove a sound" feature in Adobe Soundbooth also use spectrographic manipulation in the graphic realm as an audio restoration tool.

Paleospectrophony differs from these approaches by seeking out and playing older inscriptions that were originally conceived as graphs of time versus frequency.

1.3—Sound spectrograms in the modern sense date back only to the 1940s, when Bell Laboratories pioneered the technology for making them. That's not very long ago; we have plenty of more conventional sound recordings that are much older than that.  But it's still worth playing back a few of the earliest published sound spectrograms using reverse Fourier analysis as a proof of concept.  Since the three examples shown below are black-on-white, they had to be inverted to white-on-black before playback.

"Joe took father's shoe bench out."
Originally published in the July 1946 issue of the Journal of the Acoustical Society of America, p. 24, the spectrogram of this standard telephonic test phrase shown here was described as "made with the first laboratory model assembled from available equipment."  The frequency scale is linear and runs from 100-4000 Hz.

"You will make that line send."
This was likewise published in the July 1946 issue of the Journal of the Acoustical Society of America, p. 30, but made on a more mature version of the sound spectrograph with an analyzing filter set to a width of 90 cycles.  The frequency scale is linear and runs from 0-3500 Hz.  The gender of the speaker isn't mentioned in the accompanying text -- can you tell from looking?  From listening?

"She was waiting at my lawn."
From the book Visible Speech, published by Bell Laboratories in 1947, p. 304.  The text is given there as "she was waiting on my lawn," but "at" can plainly be heard, and "she was waiting at my lawn" was another standard telephone test phrase.   The frequency scale is logarithmic and runs from 250-7500 Hz.  This yields unusually high fidelity for a spectrogram of the 1940s. [Part 2 >>]