| With the progress of the times,wireless communication has become an indispensable technology in people’s lives.Spectrum resources are necessary for realizing wireless communication services.Therefore,the demand for spectrum resources in the entire society continues to increase.In fact,the available spectrum resources in reality are not only limited,but also non-renewable.However,the actual utilization of allocated spectrum resources is very low.Hence,to promote the development of wireless communications,it is necessary to increase the utilization of spectrum resources.Cognitive radio technology is an effective way to improve the utilization of wireless spectrum resources.Spectrum sensing technology is the precondition and key to the success of cognitive radio.Traditional energy sensing algorithms and matched filter detection algorithms are already widely known,but these methods are all proposed for stationary signal.The power spectrum of the stationary signal is independent of time,but the power spectrum of the non-stationary signal is time varying,so traditional spectrum sensing algorithms is not suitatble for non-stationary signals.Based on time-frequency analysis,this thesis studies the spectrum sensing method of nonstationary signals.The main research contents of the dissertation are as follows:1.In view of narrow-band spectrum sensing problem for non-stationary signals,this thesis studies an energy detection algorithm based on short time Fourier transform(STFT).The STFT-based energy sensing method uses the STFT to calculate the average energy of a non-stationary signal in a certain time-frequency region.Then the frequency band occupancy of the non-stationary signal in the time region is detected according to the corresponding threshold.In this dissertation,the decision threshold of the above algorithm is obtained based on the theoretical derivation results and simulation experimental data.The simulation results show that the detection performance of the algorithm is well in low signal-to-noise ratio(SNR)environment,and it is applicable not only to non-stationary signals but also to stationary signals.2.In order to improve the sensing performance of the STFT-based energy sensing method at low SNRs,this thesis studies a matched filter detection algorithm based on Wigner-Ville distribution(WVD).The matched filter detection algorithm based on Wigner-Ville distribution associates the time-frequency distribution of the signal with the time domain representation of the signal by the Moyal property of the non-stationary signal Wigner-Ville distribution.Then the occupancy of the frequency band in the time region is detected by using the matched filter sensing method.The simulation results show that the sensing performance of this method is better than the traditional matched filter detection algorithm in the case of low SNRs.3.For wideband spectrum sensing problem of non-stationary signals,this thesis combines STFT and multi-coset sampling scheme,and proposes a wideband spectrum sensing algorithm based on STFT-MUSIC and a wideband spectrum sensing algorithm based on STFT-ESPRIT.Due to the high Nyquist sampling rate of wideband signals,so the hardware consumptions is very large when the Nyquist sampling rate is used to sample a wideband signal,and it is even impossible to find an Analog to Digital Converter(ADC)that satisfies the Nyquist sampling requirements when the signal bandwidth is as high as GHz.In order to reduce the sampling rate of wideband signals,this thesis adopt the multi-coset sampling scheme,whose sampling rate is below the Nyquist sampling rate.Moreover,with the relationship between the STFT of the multi-coset sampler’s output sequences and the active sub channels,we reformulate the spectrum sensing problem as a parameter estimation problem.Then the active channel set(set of frequency bands occupied by the signal)is estimated using the Multiple Signal Classification(MUSIC)algorithm and the Estimation of Signal parameters by Rotational Invariance Techniques(ESPRIT)algorithm.The simulation results show that the proposed methods perform well with low SNR and less data samples.In addition,the developed approaches are suitable for non-stationary signals and mixed signals that contain stationary signals and non-stationary signals. |
