Energy Detection Based Spectrum Sensing for Cognitive Radios Over Time-Frequency Doubly Selective Fading Channels

Abstract

Cognitive radios may operate in practice under various adverse environments. For typical mobile and short-range scenarios, wireless links may tend to be time and frequency selective, i.e., the multipath propagations with time-varying fading coefficients will be inevitable. To cope with the encountered doubly-selective channels, in this paper we present a new spectrum sensing algorithm for distributed applications. Firstly, a dynamic discrete state-space model is established to characterize sensing process, where the occupancy state of primary band and the time-varying multipath channel are treated as two hidden states, while the summed energy is adopted as the observed output. With this new paradigm, spectrum sensing is realized by acquiring primary states and time-dependent multipath channel jointly.For the formulated problem, unfortunately, Bayesian statistical inference may be impractical due to the absence of likelihoods and involved non-stationary distributions. To remedy this problem, an iterative algorithm is further designed by resorting to sequential importance sampling techniques, thus the dynamic non-Gaussian multipath channel and primary states are estimated recursively. Another critical challenge, e.g., the noise uncertainty, is also considered, which may be incorporated conveniently into this sensing diagram and, furthermore, addressed effectively by the designed algorithm. Simulations validate the proposed algorithm. While classical schemes fail to deal with doubly-selective channels, the new sensing scheme can exploit the underlying channel memory and operate well, which provides a great promise to realistic applications.Cognotive radio, spectrum sensing, time-frequency double selective channels, dynamic state-space model, statistical inference, noise uncertainty, non-quiet sensing 1053-587X (c) 3 correlations) of primary signals [10]. In refs. [18] and [19], the probabilistic property of PUs' states is properly utilized to either design the sensing algorithm or optimize the sensing schedule strategy.It is noteworthy that, however, most existing sensing methods are tailored to a static single-path channel model. That is, wireless propagations from PUs to SUs are assumed to be invariant and, simultaneously, the multipath components (MPCs) are also ignored for simplicity. More importantly, the widespread memory distributed in time-correlated channels has been ignored unfortunately, which could be fully exploited to promote the detection performance. Therefore, the sensing performance of these methods (i.e. ED) may become less competitive in TF-DSFCs. Ref.[15] investigated the sensing performance under Rayleigh fading based on the statistical probability distribution function (PDF), which, however, can only characterize the instantaneous random behaviors of the single-path gain and may fail to track its time variations, let alone exploit the underlying channel dynamics. Besides, multipath fading will also contaminate the received signal dramatically and, therefore, results in substantial performance ...

Deep Sensing for Space-Time Doubly Selective Channels: When a Primary User Is Mobile and the Channel Is Flat Rayleigh Fading

“…The transitional behavior of time-correlated channel is modeled as a DSMC [12], [23], [25], which is shown to coincide with the dynamic nature of time-varying channel and also match existing statistical models (e.g. Clarke's model) well [23], [25].…”

“…For the non-coherent RSS observation, it is noteworthy that the channel phase will be insignificant to the DS process [12], whilst the channel gain with a specific statistic distribution, e.g. Rayleigh fading [24], is of crucial importance.…”

“…In practice, the above transitional probabilities are usually ergodic and stationary [12], [23], [26]. That means the prior densities ρ k1→k2,r are independent of the transitional time n = n/L of fading channels.…”

“…Most traditional schemes, e.g. energy detector (ED) [11], [12], unfortunately are vulnerable to information uncertainties, e.g. the large-scale spacevarying path loss or meso-scale time-varying fading channel.…”

“…the large-scale spacevarying path loss or meso-scale time-varying fading channel. As suggested by [12], [13], the fading effect, given the a priori probability density function (PDF), will be marginalized out. By focusing simply on its instantaneous random behavior, this commonly used technique, however, ignores the correlation of time-varying flat fading [13].…”