Adaptive Noise Tracking for Cognitive Radios under more realistic operation conditions

Normal operation conditions of cognitive radio applications require signal processing techniques that can be executed in real time. One of the first steps is to sense the occupied or free frequency channels. Two major drawbacks in the current techniques are that they assume (i) the noise as white and (ii) the measured spectrum as time-invariant. In real world, the noise is (i) colored so it disturbs the signal unevenly and (ii) its spectrum changes over time. Hence, tracking the time-varying noise spectrum can become crucial to remove the noise contributions and enhance the estimate of the received signal. In this paper, we study an auto-regressive model to develop an adaptive noise tracking technique using a Kalman filter such that an extension of Boll's noise subtraction technique, designed for audio noise cancellation, becomes feasible when adjusted to cognitive radio scenarios. Simulation results show the performance of this technique.