Edge based wideband sensing for cognitive radio: Algorithm and performance evaluation

Since a cognitive radio does not have fixed spectra, it may need to sense a very large frequency range to find an available band. The sensed aggregate bandwidth could be as large as several GHz. This is especially challenging if the center frequencies and bandwidths of the sensed signals are unknown and need to be detected. In this paper, an edge based wideband sensing is proposed. The method first uses the product of wavelet transforms at different scales to detect the edges (sharp changing points) of the power spectral density (PSD) of the received signal. It then forms the possible bands based on the detected edges. Thereafter, it applies a multi-band detection scheme to classify the bands as occupied or vacant. Finally, the signal to noise ratio (SNR) of each occupied band is estimated. Performance evaluation is also a complicated issue for wideband sensing. Other than the conventional metrics as probability of detection and probability of false alarm, three new criteria are proposed to evaluate the performance of a wideband sensing. Simulations are provided to verify the methods.