Design On MMSE Receiver For The Fifth Generation Wireless Modulation System

Due to the diversity of wireless communication terminals and the increment of wireless communication application requirements,higher demands have been proposed for the physical layer waveform of wireless communication.At present stage,cellular systems of the fourth generation have been optimized to provide high data rates and reliable coverage to mobile users.Cellular systems of the next generation will face more challenges,such as higher data rates,lower power consumption,lower latency and higher flexibility choice of waveforms.As a candidate of future fifth generation networks,Generalized Frequency Division Multiplexing(GFDM)has been proposed to meet the requirements above.GFDM is categorized as a non-orthogonal multicarrier technique,which suffers intercarrier interference.At present,the linear minimum mean square error(MMSE)receiver of GFDM can make a trade-off between self-interference and noise enhancement,whose bit error rate performance is not inferior to other receivers.Typically,the interference of buildings,fog and other obstacles is ubiquitous in the wireless channel,which inevitably brings fading to the received signal.Hence,frequency selection is included in the channel impulse response.However,the computation complexity of MMSE receiver is relatively high in the case of frequency selective channel,which results in high hardware cost and makes it difficult for practical application.This paper proposes two low-complexity MMSE receiver design schemes to this problem.The first one partitions the large-sized matrices involved in the receiver and explores the particular structure(sparse,quasi-tridiagonal,block-symmetric,etc.)in time domain to reduce the complexity of matrix multiplication.Meanwhile,a quasi-tridiagonal structure based 3-step matrix-inversion procedure is presented and deduced detailly,which significantly reduces the complexity of matrix inversion.On the basis of the above time domain scheme,the second scheme proposed and proves that each subblock in the partitioned matrices is a circulant matrix.Moreover,this scheme utilizes the properties(diagonalizability of the circulant matrix,low-pass pulse shaping filters,etc.)of matrices in frequency domain,resulting in further sparsity and lower computational complexity.Quantitative analysis on the complex multiplications consumed by the proposed receiver schemes is presented,demonstrating that the overall complexity of the time-domain scheme is 2 orders of magnitude simpler than the direct one,and the frequency-domain scheme achieves 0.5 order of magnitude reduction compared with existing solution that are known to have the lowest complexity.Furthermore,the simulation experiment of GFDM shows that both of the schemes do not incur any bit error rate performance loss.