3 ms·

IIRC there have been a number of papers on using a similar technique with speech to text applications.