Well, it's not really very difficult, I think.
First thing to do would probably be to dissolve the coke in warm water. Then filter to get rid of any insoluble impurites. Then add an equimolar amount of base, NH3 or NaOH both sound fine imo. This will convert the cocaine(and some other stuff) into the freebase form. Freebasae coke is not soluble in water, but from what I understand amphetamine bases and the like are. So there would be some cocaine freebase precipitate. Then one would extract the freebase from the water with ether, acetone, or a similar solvent. Anhydrous, if possible, for highest yield. Separate the two layers, and discard the aq. phase(maybe save it for a bit just to make sure you dont lose any product). Dropwise addition of HCl until the solution is just acidic will convert the cocaine back into the salt form, which is insoluble in non-polar solvents and will precipitate from the solution. Small scale recrystallization techniques should give some sexy looking crystals, which can than be filtered out of the solution, and washed with cold non-polar solvent. MP analysis would be enough to satisfy me, but you could go the extra mile and do IR or GC or TLC or w/e to make sure you got the goodies.
The only thing that worries me about this process is the step where you convert the coke back to the salt form. It has such a high solubility in water, something like almost 2g/ml I think, so wouldn't any concentration of aqueous acid result in the product being dissolved in the aqueous layer, no matter how small it is? Only thing I can thing of would be bubbling pure HCl gas through the solution, but that seems like it would be considerably more preparation/work than the other way.