GA-BFO Based Signal Reconstruction for
Compressive Sensing
Dan Li
1
, Muyu Li
2
, Yi Shen
1
Member IEEE,Yan Wang
1
, Qiang Wang
1
Member IEEE,
1 Department of Control Science and Engineering, Harbin Institute of Technology, Harbin, China
2 Department of Electronics Science and Technology, University of Science and Technology of China, Hefei, China
Phone: 86-0451-86413411; Fax: 86-0451-86413418; Email:lidanhit@163.com
Abstract—The theory of compressive sensing (CS) mainly
includes three aspects, i.e., sparse representation, uncorrelated
sampling, and signal reconstruction, in which signal reconstruc-
tion serve as the core of CS. The constraint of signal sparsity
can be implemented by l
0
norm minimization, which is an NP-
hard problem that requires exhaustively listing all possibilities of
the original signal and is difficult to achieve by the traditional
algorithm. This paper proposes a signal reconstruction algorithm
based on intelligent optimization algorithm which combines
genetic algorithm (GA) and Bacteria Foraging Optimization
(BFO) algorithm. This method can find the global optimal
solution by genetic and evolutionary operation to the group,
which can solve l
0
norm minimization directly. It has been proved
through numerical simulations that the theoretical optimization
performance can be achieved and the result is superior to that
of OMP algorithm.
Index Terms—Compressive sensing, Signal reconstruction, Ge-
netic algorithm, Bacteria foraging optimization.
I. INTRODUCTION
Conventional signal is sampled at the limit of Nyquist cri-
terion. According to Nyquist sampling theorem, the sampling
rate must be at least double higher than the highest frequency,
when reconstructing accurately the original signal. However,
with the amount of information increases, the signal bandwidth
is becoming more and more wide, which undoubtedly demands
the sampling rate and signal processing speed higher.
During the last few years, a new area called as compressive
sensing [1] [2] proposed by Dandes and Donoho solves these
problems successfully. The theory of CS mainly includes
three aspects: sparse representation, uncorrelated sampling,
and signal reconstruction [3∼5]. The reconstruction algorithm
is the core of CS theory, which is the process that the signal
is sampled by the compression and recovered to the original
high-dimensional signal. Since the dimension of the sampling
data is much smaller than the dimension of the original
signal, the signal recovery requires solving under determined
equations which may contain infinite solutions. When the
signal is sparse or compressible, the signal recovery can be
obtained by solving a constrained optimization problem of
sparse.
The constraint of signal sparsity can be implemented by the
l
0
norm minimization. However, the direct solution of l
0
norm
constraint coefficients is an NP-hard problem that requires
exhaustively listing all possibilities of the original signal,
which is difficult to achieve by the traditional algorithm. In
order to get the approximation reconstruction of the signal,
the reconstruction of the existing algorithms are converted
to the l
k
(1 ≤ k ≤ 2) norm minimization which can be
obtained by using a number of sub-optimal solution algorithm
[6∼8]: l
1
norm minimization algorithm (BP), matching pursuit
algorithm (MP), Bayesian methods (Bayesian compressive
sensing) and so on.
Intelligent optimization algorithm [9∼12] is a class of
optimization methods established by simulating a natural
imagination or process. The concept of genetic algorithms [13]
was proposed by Holland in 1973, inspired by the biological
evolution. It is a random search algorithm by simulating
natural selection and genetic mechanisms. The local searching
ability of genetic algorithm is poor, resulting simple genetic
algorithm is time-consuming, and so the searching ability is
low in late evolution.
Bacteria Foraging Optimization Algorithm (BFOA)
[14∼16], proposed by Passino in 2002, is a new comer to
the family of nature-inspired optimization algorithms. BFO
is a relatively new swarm intelligence algorithm inspired by
the foraging behavior of Escherichia coli (E.coli) in human
intestines. There exists a unique communication mechanism
between individuals of E. coli for the communication
purposes.
This paper proposed a signal reconstruction algorithm based
on intelligent optimization algorithm which combines the ge-
netic algorithm and bacterial foraging optimization algorithm
(GA-BFO) together. It can find the global optimal solutions by
continuous learning, which can solve the multimodal optimiza-
tion problems. The method proposed in this paper can solve
norm minimization directly and can reconstruct the original
signal accurately. The simulations prove that the theoretical
reconstruction performance can be achieved.
The remainders of this paper are organized as follow:
Section II will present principles of CS, GA and BFO. In
Section III, a signal reconstruction algorithm using GA-BFO
will be presented. Simulation results are given in Section IV
to illustrate the performance of the proposed method. The
978-1-4799-1334-3/13/$31.00 ©2013 IEEE
Proceeding of the IEEE
International Conference on Information and Automation
Yinchuan, China, August 2013