An increasing number of Web sites ask users to register, requesting both a user name and password lead to authentication problem when some individuals have too many passwords to be remembered. In this paper, we proposed an authentication system that used voice to identify a user. This system used Mel Frequency Cepstrum Coefficients to capture and analyse the voice signal to extract the features. Then, the Vector Quantization is implemented for feature matching to identify the unknown speaker by comparing unknown speech input with an acoustic vector in the database of voice recognition. The testing has been done with a number of users. The result show that the accuracy of this system is about 80%. This system can be use to eliminates the problem related with ID pin authentication system.
|
introduction |
feature extraction - Mel Frequency Co-efficients Cepstrum
feature matching - Cross Correlation |
OBJECTIVE |
RESULTS |