Sort:  

92% actually should be sufficient  enough to distinguish voices , but well, they should strive more to perfection.