Subsampling Based Research On Detection Technologies Of Wideband Spectrum Sensing

With the popularization of wireless devices,on one hand,the shotage of spectral source is getting more and more serious and the spectrum which can be used by unlicensed user is rare;on the other hand,the licensed frequencies are not always occupied,which results in the sparsity of spectrum.To radically resolve the problem of spectral crowding and waste and dynamically provide spectrum source to unlicensed user in real time,while not affecting the using of the licensed user,is in the scope of Cognitive radio(CR).In some particular cases that the signal of interest is weak and buried in the noise,the acquisition of this kind of signal is in the scope of signal interception(SI).Therefore,the blind detection of(weak)signal is the common problem that CR and SI both need to solve.In practice,the sampling and detection of wideband signal are the key technics and also the bottleneck of blind wideband sensing.To sample a signal of GHz band,the Nyquist rate is 2GS/s.Anolog to digital converter(ADC)supporting this rate is not only expensive but also low-resolution and power hungry.In most cases,the detection of wideband signal is blind.The existing detecting methods either suffer in low SNRs(energy detection,ED)or have high computational complexity(cyclostationary detection,CD).Considering that the wideband signal is generally sparse,to solve the above proplems,the following schemes are provided in this paper.1.Sampling scheme: we design a sampling frame that can use the available and low-cost AD board and can work at rate a bit higher than the Landau rate.The sampling channels are distinctly shifted.This scheme is similar with multi-coset sampling.2.Detetion method: by exploiting the random property of noise and interferences and using the bandwith property of communication signal,we construct frequency locator polynomials(FLPs)to locate the occupied frequencies.Design algorithm based on FLP to fully recovery the support of the original wideband signal.The contributions of this paper include the following aspects:1.Big Band,which can sense and decode wide band,is generalized and realized.2.We propose to model the problem locating occupied frequencies as a specially kind of error correction decoding.In this model,the occupied frequencies are regarded as ‘error'.The concept ‘frequency locator polynomial'(FLP),is proposed.The roots of the FLPs are one-to-one corresponding to the occupied frequencies.Based on this,we redesign the aliased spectrum resolution algorithm for Big Band.The new method improves the computing speed by nearly two orders.3.We design a sparse subsampled wideband spectrum recovery algorithm,namely AD-Big Band.In this algorithm,the bandwidth property is used.Moreover,the time shifts selection method is provided together with the related theoretical bound.The robustness is then improved.4.We propose a Nyquist sampling based blind detection algorithm for weak signal.We define a new measurement,which is similar with that of matched filter(MF).The measurement is proved to have good statistical properties.Simulation results show that the algorithm can blindly detect the signal of 1MHz bandwidth even in SNR of-27 d B,which is just 20 d B higher than the matched filter,which even requires full knowledge of the signal.5.A novel signal detection algorithm based on FLP is proposed.This method can be considered as a new kind of detecting methods beyond the popularly used methods including matched filter(MF),energy detection(ED)and cyclostationary detection(CD).The proposed detector use the FLP evaluation as a new measurement without recoverying the power spectrum or cyclic spectrum.The algorithm is proved to have strong robustness to noise.Moreover,it has lower complexity of both samples and computation compared with compressive sensing based methods.6.A new sampling method is proposed.The method combines multi-coset sampling and multi-rate sampling.In practice,a multi-channel AD board supporting external clock and several sampling rates is used.The sampling rates of all channels at any time are identical,however,this rate periodically changed between certain rates.The detection algorithm based on this sampling method is proved to efficiently lower the false alarm probability and detect the wide multiband signal even in SNR of-20 d B.