Group D - Voice recognition edit

Christian B. Olsen, Jacob H. Svalastoga, Michael J. Moltke and Thomas S. Olesen

Paper Abstract edit

It has always been an issue in different situations to be able to identify an individual. Today when more and more interactions take place electronically it becomes increasingly relevant to be able to identify an individual in this way. The ultimate way to do this is by using biometrics. In this paper we will concentrate on the specific field in biometrics called speaker recognition, identifying a person from the sound of their voice. We will extract the easily found amplitude values per timeperiod from a normal voicesample soundwave. With the Fast Fourier Transform theorem we move from the time domain to the frequency domain, and by using the Nearest Neighbour data mining algorithm we are finally able to distinquish the voice of one individual from another.

Paper and Source Code edit

  • Paper and source code is no longer online
    • Vsis-project-group-d.pdf - The paper
    • FFTLib CSharpCode.rar - Source code for sending sound to the JOP
    • JopServerWithFourier.rar - Source code for the JOP server

Previous project idea edit

Our main scope is to provide a dependable real-time distance calculation using laser equipment. In the end we hope to mount this equipment on a remote-controlled vehicle and make it able to brake at specified ranges from obstacles. The software will be developed in Java so it can run on a JOP. Project was abandoned due to the cost involved of acquiring a laser that could meet the requirements.