Public Lab Research note

spectrum matching

by jeffreyjoy | January 13, 2012 23:07 13 Jan 23:07 | #687 | #687

RE: Spectrometer kit and the using public data to perform identification of images produced.

As a general observation, it seems to me that assembling a spectrum library and matching should be similar to the way Soundhound can ID audio clips by distinctive peak matching.

Google translate uses existing wealth of translated documents upon which to base its translations. Mining the existing spectrographs from scientific journals and industrial data via Google might be something that Google Labs may be interested in participating in to support this project.

Once public databases of spectrographs are located, then this may be sufficiently interesting to convince Soundhound to allow use of its algorithm for this project.

In both cases, the commercial entities may be willing to assist because this has the potential of being a whole new market area for each of them, even though it is presently only philanthropic.


its always fascinating to me how visual and audio waveform data respond to similar technical approaches...

I'm working on this color classifier:

where i've done a very very simple n-space cartesian distance search... it looks for each color, how different it is between 2 samples, and sums the differences.

it's designed for RGB or infrared+RGB images now, but the basic approach could be expanded for as many colors (or, wavelengths) as needed. I tried to write it in a modular way so that other, better matching algorithms can be swapped in.

BTW, i tagged your post "spectrometer" so it gets filed with other spectrometry research notes; hope you don't mind.

Reply to this comment...

Good idea. This sounds pretty much like the way the search for pot plants with aerial photos. It seems to work pretty well for them.

Reply to this comment...

Login to comment.