So to put it simply: identify the frequency tinnitus is at and put a notch filter on it when listening to music?
This is definitely something I want to look into, I've never heard about it before.
Yes, that's it. A bit over-simplified, but that's it.
Some of the research papers deal with identifying not only the frequency, but the level of the tones. It gets complicated because some people have multiple tinnitus frequencies, and almost everyone has different tones in different ears. The diagnostic is tone frequency and level matching, with some very low level tones in a very quiet environment.
What I've never been able to define is the notch characteristics, though. How deep, how wide, etc. I made some assumptions based on the stability of the tinnitus tone frequency, which is that it's a fairly narrow notch, much less than 1/3 octave, and that depth beyond 20dB probably doesn't matter too much. But that's my own interpretation. What's also not well defined is the exposure level in SPL of the conditioning signals. That's fairly important, and fairly vague. Exposure time is pretty well defined, and signal content seems non-critical so long as it populates the spectrum around the notches. Primary testing was done with noise signals, and the commercial products use noise signals that are not only notched but frequency contoured as well (also not well documented, but there's a complex analysis procedure that teases out the contour). However, there's been success with notched music, and recorded sound like surf. The success rate is artificially skewed high because of better patient compliance...they like to listen to 20 minutes a day of music and surf better than pink noise, go figure.
