Using Speaker and Glossary Identification for Access Control

Abstract

Many embedded technique systems are rapidly development and integrated into people's lives, and a convenient way, voice, for human to communication with their digital embedded systems is more important today. The first thing we need to make the system recognize the user as well as the instructions issued, this will used the so-called language and vocabulary identification technology. This paper uses Gaussian Mixture Models to build a speaker recognition system with language-free and text-free features, and further utilities the Hidden Markov Models to vocabulary identification. The another feature of this system is software combined with hardware schemas, when the recognizing work of the user and vocabulary is finish, this system will sends a trigger signal to a mechanical hardware to complete the specified work. The testing environment is to simulate the general environment with some noises, so it can imitate some voice commands utility environment under various situations, and emerge the influence of the background noise for this recognition system.