Nonlinear Quantization Effects Processing And Wireless Transmission Key Technology With Low Complexity

The official issuance of the 5th-generation(5G)mobile communication system commercial license marks the official arrival of 5G era.It is important for us to consider the hardware cost and system computational complexity in 5G mobile communication systems,including but not limited to the various application scenarios and key techniques.In order to achieve the low overhead and low complexity of the next generation mobile communication systems,the nonlinear quantization effects in large systems are studied in this paper.That is,the low-resolution quantization and clipping of the orthogonal frequency division multiplexing(OFDM)system,and the mixed analog-to-digital converter(ADC)massive multiple-input multiple-output(MIMO)of massive connectivity in the scenario of massive machine-type communications(mMTC).From the perspective of compressed sensing(CS),a series of algorithms with relatively low computational complexity are proposed and extended to the above mentioned nonlinear quantization scenarios.The corresponding performance analyses validate the superiority of the proposed schemes.Firstly,based on the principle of Turbo iteration and the factor graph of message passing algorithms,the generalized Turbo(GTurbo)algorithm is proposed.The GTurbo algorithm can reconstruct the original signal in the linear measurements and nonlinear quantization based on the CS,respectively.The implementation of previous CS algorithms require many matrix multiplications with high computational complexity.Considering a special sensing matrix that allows a fast multiplication with low overhead and low computational complexity is of great interest in the large system regime.The partial discrete Fourier transform(DFT)sensing matrix,i.e.,a randomly selected DFT matrix,is one example.Moreover,low-resolution measurements of the received signal can further reduce the hardware cost.Similar to the theorem of Turbo decoding,this algorithm iteratively computes the extrinsic information of different modules,thus obtaining the final convergence value.The behavior of the GTurbo algorithm in the linear and nonlinear measurements are described by a set of state evolution(SE)equations,which are commonly used in the message passing algorithms.The numerical results demonstrate that the GTurbo algorithm can converge only in a few iterations,and the SE analysis precisely predicts the per iteration performance.Furthermore,the GTurbo algorithm and the corresponding SE analysis,can also apply to other priors and other nonlinear measurements in the regime of large systems.Consequently,this proposed algorithm can achieve system performance with low hardware overhead and low computational complexity when compared to other CS signal recovery algorithms.Then,the signal detection in the low-resolution quantized OFDM systems is investigated.Firstly,the signal detection is performed when the channel state information(CSI)is perfectly known.Secondly,the signals are detected with pilot-aided channel estimation.The GTurbo algorithm can be directly applied to detect the quadrature amplitude modulation(QAM)signals under the case with perfect known CSI.Similarly,the performance analysis of QAM signal detection is obtained by deriving a set of SE equations.As expected,the curves of the simulation results and the theoretical results match very well.When it comes to the pilotaided channel estimation,a Bayesian posterior minimum mean square error(MMSE)can be applied here.The elements of OFDM frequency channel can be modeled as Gaussian distribution when there are many channel taps in the time domain.Further,the QAM signals are detected based on the estimated CSI in the framework of GTurbo algorithm.The performance gap between different CSI is very small due to the accuracy of the Bayesian posterior MMSE channel estimation and the superiority of the GTurbo-architecture-based signal detection algorithm.Next,the high peak-to-average power ratio(PAPR)and the clipping method with low complexity in the OFDM systems are studied.The expectation maximization(EM)-Gaussian mixture(GM)-GTurbo algorithm is proposed for clipping estimation by using reliably detected subcarriers to solve the problem of the nonlinear distortion.In order to increase the bandwidth efficiency,the location-based selection and Turbo-decodingbased selection are provided to identify reliable subcarriers in non-coded OFDM systems and coded OFDM systems,respectively.These reliable subcarriers act as the pilots in the recovery of clipping signals.The elements of the clipping signal can be modeled as the GM distribution,where the prior knowledge is learned through the EM algorithm.The performance analysis of the EM-GM-GTurbo algorithm is obtained by the equations of SE.Numerical experiment results demonstrate that the proposed technique achieves a significant performance improvement in terms of signal recovery accuracy,i.e.,bit error rate(BER),and the agreement of the theoretical analysis with SE,i.e.,mean square error(MSE).The final value of the MSE obtained iteratively by SE agrees well with that acquired through the EM-GM-GTurbo algorithm.The convergence rate of this scheme is very fast and the computational complexity is relatively low when compared to other nonlinear compensation algorithms.Finally,the channel estimation and active user detection are investigated for massive connectivity in the scenario of mMTC.A partial DFT pilot sequence is provided to assist joint channel estimation and user activity detection scheme for massive connectivity,in which a large number of devices with sporadic transmission communicate with a base station(BS)in the uplink.The joint channel estimation and device detection problem can be formulated as a CS single measurement vector or multiple measurement vector(MMV)problem depending on whether the BS is equipped with single or large number of antennas.Due to the high hardware cost and power consumption in massive MIMO systems,a mixed ADC architecture is considered.The proposed GTurbo-MMV algorithm can precisely estimate the CSI and detect active devices with relatively low overhead.Furthermore,the SE is studied for the MMV problem to obtain achievable bounds on channel estimation and device detection performance,in which both the missing and false detection probabilities can be made tend to zero in the massive MIMO regime.The simulation results confirm the theoretical accuracy of our analysis.