Robust blind spectral estimation in the presence of non-Gaussian noise.

Dynamic spectrum access (DSA) has seen a growing interest in recent years due to spectrum scarcity. Cognitive radio is a necessary component to enable DSA. One key requirement for cognitive radio is the ability to blindly detect the presence of signals in time and frequency. One popular method for blind detection is computing a power spectral density estimate using the Welch periodogram. The Welch periodogram is optimally efficient under the assumption that received signals are in the presence of additive white Gaussian noise. However, the Welch periodogram is not robust against non-Gaussian noise. Robustness against non-Gaussian noise is desirable because many man-made devices-such as microwave ovens, WiFi co-channel, gas emissions, etc.-contaminate the spectrum with non-Gaussian noise. In this work, robust estimation techniques are applied to the Welch method to increase its robustness against non-Gaussian noise at the cost of estimation efficiency and computational ease. The increase in robustness of the modified Welch method (compared to the traditional Welch method) is demonstrated for the Middleton Class A non-Gaussian noise model. These robust techniques are shown to enable more reliable detection of signals contaminated by non-Gaussian noise.