Distributed spectrum sensing via energy detection in cognitive radios

Energy Detection is widely studied spectrum sensing system to detect signal energy due to its simplicity and lack of requirement signal characteristics. This thesis focuses on two main problems on distributed detection of energy detectors. The first problem is threshold finding in the nodes due to achieve constant false alarm probability ($P_F$) for binary decisions. The second problem is to derive optimum quantization of node observations due to minimum distortion and to give a expression on number of bits should nodes quantize their observations. For adaptive threshold finding, two algorithms are proposed. The first algorithm depends on minimizing the distance between instantenous $P_F$ and desired $P_F$ by Netwon's method. This algorithm is implemented for both mathematically calculated and estimated $P_F$ values. The second algorithm depends on generating a discrete random variable and comparing it with the previous decision in every iteration. By this algorithm, system adjusts its threshold to any changes in the noise variance and works totally independent from $P_F(\lambda)$. In the second focus of the thesis, minimum distortion quantization of two hypothesis together is derived for Lloyd-Max algorithm and its performance is observed. Number of quantized bits of observations is also in the scope of this thesis. A lower bound on rate-distortion function is derived for the signal which consists of two hypotheses. Another approach on number of bits is handled by Chernoff bound which allows to compare performance of different bit quantizatons.