For me it sounds like ethnic vocals treated with some automated fx.
I would try to get a great sustained vocal/speech phrase and chop it up to get some rhythm. Then i would use celemonys melodyne to pitch individual notes around. In melodyne you can change the formants of the vocal without editing the pitch to get the "kittiness" into the vocal. Also be sure to use the pitch drift function of melodyne to get some pitch shifts, also reduce the vibrato with the pitch modulation tool to make the sounds less human and more instrumental.
A good idea is to pitch stuff up and the formants down or pitch stuff down and the formants up. that way it stays vocal but changes timbre. Also remove noisy parts of the vocal like s, f p, t and such syllables as they tend to screw melodynes algorithm. Vowels like a,e,i,o,u work best.
Then i also hear some kind of filter. You hear they there is some lowpassfilter. I would automate the cut-off according to the phrase you wrote. Try to accent another frequency area with a notch filter sweep according to the phrase (it is like following the natural pitch movement). The area between 700 and 900 hz should work fine, since there is not that much information in our speech.
Also distort the voice slightly and compress the hell out of it to get a stable instrument. Flangers phasers and other modulation effects can give some great movement. Just copy your phrase 100 times and then adjust slow modulation parameters. then render it to audio and search for the best peaces to rebuild one great phrase.
Another way to maybe achieve it is to put the vocal chops into a sampler and pitch/pitchshift it there. most samplers have inbuilt filters so you can have everything in one package!
I hope this gives you a start.