Musical password based biometric authentication
Ravi Prakash, Suresh Kumar V, Chandan Kumar, K. K. Mishra · 2016
Human voice has some unique features which enables us to recognize the person whose voice has been previously heard by us. Voice as a bio-metric feature has got a lot attention for person authentication, identification. A voice bio-metric system works on the human voice model and the features of vocal cord and tracts. In this paper we propose a system which authenticates the user by recognizing the musical pattern sung by the user as a password. The sung musical sequence is first segmented at note level; the musical notes are then classified into the 12 chromatic musical notes. The notes with their duration are then used by the system at enrollment phase to generate the template to be stored as the password. At the authentication phase the musical sequence is again segmented and classified and the password is reconstructed which is being matched with the pre-stored template. The voice recognition system also uses the user level features for authenticating the legitimate user. The proposed algorithm has been implemented and tested on a relatively small data set containing 20 users with 10 male and 10 female users. The overall recognition accuracy of the system has been found to be 75 per cent.