Speaker recognition systems are useful for deriving intelligence information by tracking telephone conversations of people who are perceived to be a threat to the national security. Rules prohibit security agencies from eavesdropping on telephone conversations of private citizens. However, it may be necessary to eavesdrop on people with questionable national integrity. A speaker recognition system will assist the security agencies to filter out only the necessary telephone calls to be eavesdropped on, giving reasons that can be justified according to the law of the land. This project proposes to develop a text/language independent speaker recognition system.
Several laboratory experiments only require sub-Mega sampling rates. Multiplexing frontend for the generator / analyzer hardware will enable interfacing of multiple measurement and control interfaces for such applications. Such an interface is very helpful to reduce the cost involved in setting up multiple virtual lab equipment scaling up all over the country to support huge amount of users.
Multi-user Shared Environment will enable multiple users to login to the system and perform experiments without the user feeling any latency involved with multiplexing of hardware devices. The amount of simultaneous users supported in the system will depend on various experiment related constraints, such as mode of experiment (batch / interactive), time constant for performing the experiment, and so on.
