Identification of push-to-talk transmitters using wavelets and spectral correlation

Abstract

The purpose of this thesis is to find an automated way to identify push-to-talk transmitters using a wavelet or a spectral correlation based approach. In the wavelet approach, a distance algorithm is applied to the wavelet scales of the signal and the template. One signal from each transmitter set is taken as a template. The distance algorithm computes the distance between the local extrema of the wavelet coefficients of the template and the signal. Results show that the Wavelet Transform (WT) distance algorithm is able to classify the four signal sets accurately. Good identification results are achieved even at low signal-to-nose ratios. In the spectral correlation approach an averaged template for each signal is used. The spectral coefficients for templates and signals are computed by extracting the magnitude squared of the Fast Fourier Transform (FFT) of the data. This method performs better for most signals better than the wavelet method because it can identify at lower signal to noise levels than the wavelet method does.