Training Sequence Design for Efficient Channel Estimation in MIMO-FBMC Systems

Training Sequence Design for Efficient ChannelEstimation in MIMO-FBMC Systems



This paper is focused on training sequence design forefficient channel estimation in multiple-input multiple-output filterbankmulticarrier (MIMO-FBMC) communications using offsetquadrature amplitude modulation (OQAM). MIMO-FBMCis a promising technique to achieve high spectrum efficiency aswell as strong robustness against dispersive channels due to itsfeature of time-frequency localization. A salient drawback ofFBMC/OQAM signals is that only real-field orthogonality canbe kept, leading to the intrinsic imaginary interference beinga barrier for high-performance channel estimations. Also, conventionalchannel estimations in MIMO-FBMC systems mostlysuffer from high training overhead especially for large number oftransmit antennas. Motivated by these problems, in this paper, wepropose a new class of training sequences which are formed byconcatenation of two identical zero-correlation zone sequenceswhose auto- and cross- correlations are zero within a timeshiftwindow around the in-phase position. Since only realvaluedsymbols can be transmitted in MIMO-FBMC systems,we propose “complex training sequence decomposition (CTSD)”to facilitate the reconstruction of the complex-field orthogonalityof MIMO-FBMC signals. Our simulations validate that theproposed CTSD is an efficient channel estimation approach forpractical preamble-based MIMO-FBMC systems.


A novel preamble design approach based on complex training sequences decomposition (CTSD) to facilitate the reconstruction of complex-field orthogonality in MIMO-FBMC systems. We propose to cascade two identical ZCZ sequences in the time domain forming preambles with non-zero values in odd or even subcarriers only (in the frequency domain). By doing so, ICI can be mostly self-cancelled as it is mainly from the neighboring subcarriers. Nevertheless, these

newly designed training sequences, which are complexvalued in the frequency domain, cannot be directly applied to MIMO-FBMC systems because only real-valued transmitted symbols are allowed.


Orthogonal frequency division multiplexing (OFDM) is anefficient multicarrier modulation scheme which is resilientto the effect of multipath fading channels. Although all theOFDM subcarriers are modulated by waveforms that arelimited in the time-domain, in practice, there is unavoidablepower leakage in the frequency-domain and because of this,the guard band has to be placed so as to minimize adjacentchannel interference to other coexisting wireless systems  Furthermore, OFDM’s robustness against multipath propagationrelies on the insertion of cyclic prefix (CP) which is a lossof spectrum efficiency.

As an alternative modulation scheme to CP-OFDMsystems, this paper focuses on the filterbank multicarrier(FBMC) systems employing offset quadrature amplitude

modulation (OQAM) without inserting CP, called FBMC/OQAM . For ease of presentation, from nowon, we use FBMC to denote FBMC/OQAM. The subcarriersin an FBMC system are modulated with staggered OQAMsymbols.


  1. Du, “Pulse shape adaptation and channel estimation in generalizedfrequency division multiplexing systems,” School Elect. Eng., Licentiatethesis, Roy. Inst. Technol., Stockholm, Sweden, Dec. 2008.
  • Based on the assumption of low frequency selectivity, interference approximation method (IAM) which is capable of computing an approximation of the interference from neighboring symbols.
  1. Hu, G. Wu, T. Li, Y. Xiao, and S. Li, “Preamble design with ICIcancellation for channel estimation in OFDM/OQAM system,” IEICETrans. Commun., vol. E93-B, no. 1, pp. 211fi214, Jan. 2010.
  • Observing ICI mainly from the nearest subcarriers, another ICM was developed by constructing preambles with only odd- (or even-) indexed subcarriers.


  • Complexvalued in the frequency domain, cannot be directly applied toMIMO-FBMC systems.
  • Only real-valued transmitted symbols are allowed.


in this paper, we proposea new class of training sequences, which are formed by concatenation of two identical zero-correlationzone sequences whose auto-correlation and cross correlation are zero within a time-shift window around the

in-phase position. Since only real-valued symbols can be transmitted in MIMO-FBMC systems, we propose“complex training sequence decomposition (CTSD)” to facilitate the reconstruction of the complex-fieldorthogonality of MIMO-FBMC signals. Our simulations validate that the proposed CTSD is an efficientchannel estimation approach for practical preamble-based MIMO-FBMC systems.


Training Sequence Design for Efficient ChannelEstimation



As shown in Fig. 1, an equivalent FBMC baseband modelwith M subcarriers is considered, in which the subcarrierspacing is 1=T , with T being the complex symbol interval.The transmitted symbol am;n is real-valued with frequencyindex m and time index n, and T =2 is the interval of realvaluedsymbols. am;n and am;2nC1 are obtained by takingthe real and imaginary parts of complex-valued symbol fromquadrature amplitude modulation (QAM) constellation.2 g .l/is the employed symmetrical real-valued prototype filterimpulse response with total energy of one and with lengthof Lg.


Let [h(0); h(1); _ _ _ ; h(Lh 􀀀 1)]T be the discrete impulseresponse of a multipath fading channel, where Lh denotesthe maximum channel delay. According to (1), the basebandreceived signal therefore can be written as

where r(l) denotes the complex additive white Gaussian noisewith zero mean .


  • Efficient channel estimation in MIMO-FBMC systems.
  • Transmit them separately over two FBMC symbols only.
  • This saves huge amount of precious time-frequency resources as the resultant training overhead does not depend the number of transmit antennas.



  • System : Pentium Dual Core.
  • Hard Disk : 120 GB.
  • Monitor : 15’’ LED
  • Input Devices : Keyboard, Mouse
  • Ram : 1GB


  • Operating system : Windows 7.
  • Coding Language : MATLAB
  • Tool : MATLAB R2013A


Su Hu, Zilong Liu, Yong Liang Guan, Chuanxue Jin, Yixuan Huang, Jen-Ming Wu, “Training Sequence Design for Efficient ChannelEstimation in MIMO-FBMC Systems”, IEEE Access, 2017.


About the Author

Leave a Reply