Effects of equipment variation on speaker recognition error rates

摘要

Speaker recognition is one biometric used in identity authentication. As use of biometric systems become more widespread the need for robustness in various environments and applications becomes more significant. One such variable is recording equipment. Speaker recognition systems in distributed environments, such as the internet, are likely to use different microphones to perform identity authentication. The success rates for speaker recognition systems are dependent upon the variability of the recording system performance in a speaker recognition system. This paper investigates how speaker recognition false accept, false reject rates are affected by variations in recording systems.