where k is a normalizing constant whose value depends on the number of samples used, d is the duration of the answer, xn is the nth sample of the received voice response, and the primed values are the corresponding quantities taken from the reference voice record.