analyse speech recording and phonetics

Julian Skar's icon

I want to record a female voice into a buffer. then splice up the sound file AND divide the chunks into different categories from complex spectral content to simple spetral content. This way im hoping to be able to filter out interesting phonetics with interesting percussive quality.
This phonetics i will exagerate and manipulate in realtime.

I have experimented with the splice~ object and modified similar objects/patches. This works great for splicing sound but i really have a long way to go before the categorizing task is complete. I have looked at Ircams FTM Gabor objects, analyzer~ and others but i dont know what would be the best way, and it will take ages to get my head around this Any thoughts or hints would be higly appreciated.