Design and Development of Voice Authentication System

Abstract

Spoken language is the most natural way used by humans to communicate information. The newlinespeech signal conveys several types of information. From the speech production point of view, newlinethe speech signal conveys linguistic information (e.g., message and language) and user newlineinformation (e.g., emotional, regional, and physiological characteristics). From the speech newlineperception point of view, it also conveys information about the environment in which the speech newlinewas produced and transmitted. Even though this wide range of information is encoded in a newlinecomplex form into the speech signal, humans can easily decode most of the information. Such newlinehuman ability has inspired many researchers to understand speech production and perception for newlinedeveloping systems that automatically extract and process the richness of information in speech. newlineThis speech technology has found wide applications such as automatic dictation, voice command newlinecontrol, audio archive indexing and retrieval etc. newlineThe application defines which information in the speech signal is relevant. For example, the newlinelinguistic information will be relevant if the goal is to recognize the sequence of words that the newlineuser is producing. The presence of irrelevant information (like user or environment information) newlinemay actually degrade the system accuracy. In this thesis, we deal with automatic systems that newlinerecognize who is speaking (the user s identity) [22] [14]. newlineIt was Lawrence Kersta who made the first major step from user identification by humans towards newlineuser identification by computers when he developed spectrographic voice identification at Bell newlineLabs in the early 1960s. His identification procedure was based on visual comparison of the newlinespectrogram, which was generated by a complicated electro-mechanical device [17]. Although the newlinevisual comparison method cannot cope with the physical and linguistic variation in speech, his newlinework encouraged the introduction of automatic user authentication. In the following four decades, newlineuser authentication research has advanced a lot. Some commerc