Continuous speech recognition and identification of the speaker system

Abstract

Currently speech recognition and speaker identification based on a biometric parameter such as voice have been treated as two different worlds and in the market there are no integrated applications of these systems. The design of a system could mean a great contribution to the development of personalized commands, in the area of home automation and robotics, thanks to the availability of the message and the identification of the speaker. Therefore, the development of an integrated biometric voice system is proposed, based on a single voice sample for the identification of the speaker and the message. We use GOOGLE SPEECH API, as a voice text translation tool, and Mel Frequency Cepstral Coefficients or MFCCs extracted from voice signal to identify speakers voice. Functional tests were carried with 50 randomly users, in the end of the study results show 96.4% efficiency in identification, demonstrating efficiency using MFCCs in speaker’s automatic recognition and verifying the use of GOOGLE SPEECH API as a fast, accurate and robust translation tool.