FPGA Implementation for GMM-Based Speaker Identification

Table 2

Hardware and software results for testing with 5 s of test utterance.

Parameter	System
	Hardware; (XC2V6000 at 48 MHz)	Software
Data transfer	16.8 ms for 500 vector transfer	Not applicable
Classification (clock freq.  = 50 MHz)	0.8 ms per vector for speaker set of size 20	69 ms per vector for speaker set of size 20