Cognitive Sensing and Transmission Strategies
Many frequency bands for wireless services are severely underutilized by the primary users (PU) to which these bands are assigned. This motivates cognitive radios (CR), which identifies vacant spectrum and transmit accordingly. Many algorithms, some as simple as energy detection, are used to sense the spectrum. However, in the wideband sensing or, generally, low-SNR regimes, the detection performance of such algorithms is limited. To address that, a new class of statistical tests known as sequential tests have been introduced, which accumulates a test statistic until it reaches one of two thresholds. In paper A, we introduce new time-varying thresholds for sequential spectrum sensing.
These new thresholds, for an SNR of -10 dB, in comparison with standard sequential detection with parallel (fixed) thresholds with similar probabilities of mis-detection and
false-alarm, performs 54% faster in terms of maximum detection time (90 percentile). In papers B and C, we study CR transmission strategies that are based on noisy observations of the PU activities, which is modeled as a two-state discrete-output hidden Markov model (HMM). We introduce a transmission strategy which is based on comparing an a-posterior probability (APP) log-likelihood ratio (LLR) with a threshold. The objective is to maximize the utilization ratio (UR), i.e., the relative number of the PU-idle slots that are used by the CR, subject to that the interference ratio (IR), i.e., the relative number of PU-active slots that are used by the CR, is below a certain level. We demonstrate a more than 300% increase in UR over standard energy detection, for the same IR value, at the SNR of -5 dB. Finally, in paper C, we use a continuous-output HMM to model the received signal, and calculate an APP LLR based on that. We show that this strategy is the optimum in the sense of maximizing the UR, given a certain maximum allowed IR, among all CRs. Moreover, two practical schemes for calculating the transmission threshold are introduced. Numerical results show that the first method yields a threshold which is close to optimum when the PU use a large fraction of the available spectrum (i.e., when the PU activity level is high). However, the method fails to give a valid threshold for (i.e., a threshold that respect the maximum allowed IR) low SNRs when the PU activity level is low. The second method is analytically proven to always give a valid threshold, regardless of SNR and PU activity level. The resulting UR is reasonably high when the PU activity level is high, but quite low when the PU activity level is low. In addition, an upper-bound for the UR of any CR strategy is presented. Simulation results have showed over 116% improvement in UR at SNR of -3dB and IR level of 10% with PU state estimation over energy detection.
Hidden Markov Model
Sequential Spectrum Sensing