Pattern-Aware Sequential Monte Carlo Detection for
Generalized Space-Time Shift Keying*
Jianping Zheng
1
, Juan Tao
1
, Jinfang Dou
2
, Baoming Bai
1,2
1
State Key Lab of ISN, Xidian University, Xi’an, 710071, P.R.China
2
National Key Laboratory of Science and Technology on Space Microwave, Xi’an, 710000, P.R.China
Email: jpzheng@xidian.edu.cn, t377902391@126.com, jinfang_dou@163.com, bmbai@mail.xidian.edu.cn
Abstract—In the generalized space-time shift keying (G-STSK)
modulation, the constraint of the activated dispersion matrix
pattern makes the application of the sequential Monte Carlo
(SMC) detector become nontrivial. In this paper, the pattern-
aware variants of the stochastic SMC (SSMC) detector and the
deterministic SMC (DSMC) detector for the G-STSK modulation
are studied. Specifically, a pattern-aware importance sampling
procedure is first presented to take into account the constraint of
the activated dispersion matrix pattern in the SMC framework.
Then, the SSMC detector generates symbol samples based on
importance sampling and resampling techniques; whereas the
DSMC detector recursively performs exploration and selection
steps in a greedy manner. Computer simulation results are
provided to demonstrate the performance of the SMC detectors.
Keywords—space-time shift keying; sequential Monte Carlo;
soft-input soft-output
I. INTRODUCTION
Recently, a universal multiple-antenna architecture, termed
generalized space-time shift keying (G-STSK) modulation,
has been proposed in [1][2]. The G-STSK modulation is based
on the activation of P (P1) out of Q appropriately indexed
space-time dispersion matrices. To be more specific, G-STSK
maps the information to both the activated dispersion matrix
pattern and the constellation points which are used to
modulate the corresponding activated dispersion matrices. By
optimizing both the number and size of the dispersion
matrices, the number of activated dispersion matrices, as well
as the number of transmit and receive antennas, the G-STSK
modulation can provide flexible diversity-multiplexing
tradeoffs. In fact, G-STSK includes many typical multiple-
antenna architectures as its special cases such as vertical Bell
laboratories layered space-time (VBLAST) architecture [3],
linear dispersion (LD) code [4], orthogonal space-time block
code (OSTBC) [5], spatial modulation(SM) [6], and space-
time shift keying (STSK) [7][8].
In the G-STSK modulation, to enhance the bandwidth
efficiency, it is desirable to increase the value of P. However,
upon increasing the value of P, the G-STSK receiver has to
cope with the equivalent number of inter-channel interference
(ICI) contributions [2]. Thus, a higher P leads to a higher
computational complexity at the receiver. In fact, the optimal
maximum likelihood (ML) detector may be prohibitively
complex for a high-P scenario. Therefore, it is urgent to
develop efficient near-optimal detectors.
Recently, Sugiura et al [9] proposed a matched filter (MF)-
based soft-decision detector (SoD), where a reduced-
complexity approximation of the soft-decision information
metric is presented. However, its complexity is still
exponential and its performance has a medium gap from that
of the optimal detector. Although the Markov chain Monte
Carlo (MCMC) sampler [10] is also proposed in [9] to reduce
the complexity of the MF-based SoD, its complexity is still
very high. A well-known detector for the conventional multi-
antenna spatial multiplexing architecture is the sphere
decoding (SD) algorithm [11]-[13]. However, to employ the
SD detector in G-STSK, f(Q,P) SD algorithms should be
needed, where f(Q,P) is the binomial coefficient of Q and P.
On the other hand, the worst-case complexity of each SD
algorithm is also exponential [14].
In this paper, we study the efficient realization of the
sequential Monte Carlo (SMC) detector for the G-STSK
scheme. The SMC detector shows excellent detection
performance with fixed and low complexity in the
conventional multiple-antenna system [15][16]. To facilitate
the SMC detector for the G-STSK scheme, a pattern-aware
importance sampling procedure in the SMC framework is
presented to take into account the constraint of the activated
dispersion matrix pattern. Similar to [16], the SMC framework
under both stochastic and deterministic settings is considered.
And two detectors, the stochastic SMC (SSMC) detector and
the deterministic SMC (DSMC) detector, are developed.
Simulation results show that both the proposed detectors can
approach the good performance in both un-coded and coded
systems with fixed polynomial complexity.
II. G-STSK
MODULATION
A. System Model
Consider the G-STSK(N
t
, N
r
, T, Q, P) scheme with N
t
, N
r
,
T, Q and P denoting the numbers of transmit and receive
antennas, the interval per space-time codeword, the numbers
of total and activated dispersion matrices, respectively. The
input-output relation can be represented by
YHSV
, (1)
where
r
NTu
Y
is the received signal over the N
r
receive
antennas. The entries of the channel matrix
rt
NNu
H and the
noise matrix
r
NTu
V
are assumed to be circular symmetric
This work was supported in part
y the National Basic Research Program
of China (973 Program, No.2012CB316100), the National Natural Science
Foundation of China (No: 61201140, 61372074), and National Key
Laboratory Foundation of China (No: 9140C530401120C53201).
2014 IEEE International Conference on Computer and Information Technology
978-1-4799-6239-6/14 $31.00 © 2014 IEEE
DOI 10.1109/CIT.2014.59
36