Basic Project Description:
The aim of this project was to implement Speaker Verification.
This has been realised, by creating a client-server application where the user can use the software from a client machine.
The user is given a random piece of text to read into a microphone. This data is then sent to the server machine, where the software
will decide if this person is who they claim to be.
In order to identify different users, word model's for each user in the system were built, based on 17 training utterances.
These training utterances covered all grammar components fairly. Once these 17 WAV files were recorded and sent to the server,
a perl script was called to create the user independent model.
The next step is to record 5 testing utterances, upon which a user's individual threshold is calculated. These 5 WAV's
are recorded and sent to the server, where another perl script moves, transforms and compares them to existing user models
and set a personal threshold in file stored in the user's directory on the server.
Once a model has been created for a user and their independent threshold set, the user can logon to the system.
Logon is carried out by receiving a randomly generate sequence of nine words from the server, recording the utterance,
sending it to the server and calculating a cumulative log-likelihood from the MLF (Master Label File) generated. This figure
is then compared to the individual threshold and a decision to Accept/Reject the identity claim is taken.
In our program, the X: Drive is then mapped to a "secure" folder on the server, as an example to an application of this
Project. In place of this, any other code could be placed: Sending file(s) from Server to Client, Printing Special
Information to the User's screen or opening applications/URL's. Any information that would normally be behind a password
or PIN Number, would be well suited to this project.
© Ronan Crowley 2001
© Paul Connolly 2001