ThinkIT's Speaker Identification Engine (TSIE)

By Beijing Kexin Technology Co.

Introduction:

Our speaker identification engine (TSIE) integrated our leading speech-signal processing technology with statistic modeling technology, it's capable to identify a speaker just by analyzing a short voice message. This engine can meet the special needs purposed for web safety, state security and speaker identification, etc..

Operating Environment and Requirement:

Sounds input device (internal microphone);

PC or server with CPU 1.6G or higher (Intel Pantium4 or equivalent);

Operating system: Windows 2000 or equivalent OS;

Main memory: 256Mb or higher;

Key feature:

Besides the applications as mentioned above, in other applications, this system can also perform in high recognition accuracy.

Limited training voice and flexible test speech length make it naturally in real usage mode.

Performance and speed :

For speaker identification application, one standard PC can support 8 incoming calls at the same time;

Within 200 persons, recognition accuracy can reach 99% (15 seconds registration, 6-10 seconds identifying);

Confidence measure, which can effectively deny any unregistered speaker.

Programming mode and porting:

Our TSIE is sealed in a DLL file, it provides a group of C type API functions which can be used for other applications. The application procedure works in a principle that the system communicates with the engine via processing callback functions.