This thesis proposes solution for two different problems. One it provide solution to extract the frequency band boundaries information efficiently from the observed spectrum. Second it performs spectrum sensing on the signal acquired at less than Nyquist rate samples. Applying wavelet transform on the observed wideband spectrum generates edges (peaks) which contain information regarding frequency band locations. In the presence of noise, the wavelet transform generates a mixture of true peaks and noisy peaks. A threshold value is proposed for extracting the true peak information. Sensing the wideband spectrum puts constraints on hardware. Spectrum sensing is a time dependent process. Sampling a wideband signal based on Nyquist sampling theorem may require more time than given sensing duration. To solve this problem, observed spectrum is acquired at less than (required) Nyquist rate samples. Spectrum sensing is performed using the structure based Bayesian sparse reconstruction algorithm. Comparison of results with the techniques present in literature showed considerable improvement. |