Signals analysis is commonly applied to voice signals. Voice signals are modulated using Amplitude Modulation (AM) for broadcast distributions, and signals are compressed before transmission when speaking into a phone. This report will introduce the use of the Fourier Transform into speech signals analysis. Specifically, we will attempt to use the Fourier Transform to identify the speaker of a series of words.
To focus our efforts, we present the following problem:
Biometric identification has begun to be used to maintain security. Fingerprint identification door locks are now sold through common stores like Staples1. A development team in Spain has created an IRIS scanner that performs person verification to unlock cellular phones using the phone’s built in camera2. Many biometric identification sensors are expensive, and therefore they have yet to proliferate. A common microphone, though, is very cheap – even cheaper than a camera. This paper investigates the possibility of using voice signals to perform identification.
Specifically, we expect a user to say four numbers from the set “zero” through “nine”. These four numbers are the user’s Personal Identification Number (PIN). We will make a comparison of each stated number with previous recordings of that user, and use those comparisons to make a decision as to whether or not there is a match to the user.
