You are on page 1of 6

2015 Second International Conference on Advances in Computing and Communication Engineering

Design of FIR Filter using Biogeography Based


Optimization
Mandeep Kaur Urvinder Singh Dilpal Singh
Chandigarh University, Gharuan Chandigarh University, Gharuan Chandigarh University, Gharuan
Mohali, India Mohali, India Mohali, India
er.mandeep14@gmail.com urvinders@gmail.com dilpal.singh01@gmail.com

Abstract—Digital signal processing (DSP) systems require particle swarm optimization (IPSO) and craziness based
filters to fulfill their needs for particular frequency particle swarm optimization (CRPSO) whose results were
characteristics. This paper presents designing and optimization much improved as compared to real code genetic algorithm
of digital finite impulse response (FIR) highpass filter. (RGA), particle swarm optimization (PSO) and Park and
Optimization is a process to bring out the best result for an McClellan algorithm (PM) [11, 12, 13]. The algorithms used
objective function. BBO is the algorithm which is based on the to optimize problem function were basically adopted from
biogeography of organisms in a habitat. The algorithm has nature and its habitants. Simon proposed another algorithm
certain modified variants such as BBO with chaos, enhanced named biogeography based optimization (BBO) algorithm
BBO, blended BBO and BBO with immigration refusal (IR)
which has some unique features as compared to other biology
which are used to optimize the filter and then compared with
based algorithms. This algorithm was demonstrated on real
each other. The works concludes that out of various variants,
blended BBO shines out with optimum results for a given set of
world sensor selection problem and it proved better than
parameters. The convergence profiles of these variants proved various other algorithms chosen for comparison [6]. The
blended BBO as better converging variant of BBO. algorithm was also used for dynamic deployment of static and
mobile sensors [7]. Du et al. implemented features of other
Keywords—FIR-finite impulse response;BBO-bio-geography heuristic algorithms i.e. evolutionary strategy and immigration
based optimization; HSI- habitat suitability index;SIV-suitability refusal with BBO to modify the algorithm. The modified
index variables; IR-immigration refusal; version of BBO performed better than original BBO when
certain tests were implied over the new version BBO [8]. Ma
I. INTRODUCTION and Simon proposed BBO with blended migration for
Filters are the electronic circuits which act as building constrained optimization problems and compared the modified
blocks for all digital signal processing (DSP) system [1]. BBO with GA and PSO [9]. BBO has been used to optimize
Digital filters are the filters which give output in digital form the antenna and it provided superior results as compared to
and are faster than analog filters when compared on the basis other optimization algorithms [18-24]. In another work same
of performance, output and reliability. Digital filters can be algorithm was compared with stud genetic algorithm (SGA)
classified into two types: Finite Impulse Response (FIR) and and standard particle swarm optimization 2007 (SPSO 07)
Infinite Impulse Response (IIR). FIR filters are more stable, where it was demonstrated that BBO performed better than
free from phase distortion and less sensitive as compared to other competing algorithms [14]. Another modification was
IIR filters [3]. proposed by Pattnaik et al. where clear duplicate operator was
introduced into BBO and was tested over different benchmark
Litwin explored the basics of digital filters and difference functions. Enhanced BBO (EBBO) gave better results when
between FIR and IIR filters [1]. Further, Mastorakis et al. compared to BBO and its other variants [10]. Recently,
designed a two dimensional filter using Genetic Algorithm another modification of BBO was proposed as biogeography
(GA) and compared it with some previous designs. The results with chaos to improve the performance of BBO for multi-
showed that stability can be guaranteed in this method with objective problems and it has been proven feasible [15].
simplification of the filter [2]. Karaboga and Cetinkaya
designed digital FIR filter using differential evolution (DE) This paper presents various migration variants of BBO.
algorithm and its performance was compared with genetic The variants are enhanced BBO, blended BBO, immigration
algorithm and least squares method. It was found that DE refusal BBO and BBO with chaos. Each variant tries its best to
calculated optimal results quicker than GA [3]. Ababneh et al. find the most optimum result i.e. fitness. All four variants are
designed linear phase FIR filter with the help of particle compared on the various platforms and conclusions are
swarm optimization (PSO) and GA where PSO outperformed brought out.
GA [4]. Luitel designed linear phase FIR filter using PSO and The paper has been divided into various sections. Section I
differential evolution particle swarm optimization (DEPSO) gave a brief introduction about filters and BBO. Section II
and concluded that DEPSO proved to be a better choice for explains about FIR filter and the objective function. The
FIR filter in dynamic environment [5]. Mondal et al. designed following section i.e. section III explores the algorithm
linear phase highpass FIR filter by implementing improved

978-1-4799-1734-1/15 $31.00 © 2015 IEEE 312


DOI 10.1109/ICACCE.2015.136
implied with its variants. Further section IV discusses the III. BBO AND ITS VARIANTS
results with graphs and tables. Meanwhile section V tells Biogeography based optimization is the algorithm which is
about conclusion withdrawn. based on concept of biogeography i.e. distribution of
II. HIGHPASS FIR FILTER DESIGN organisms over space and time. The geographical area is
named as “island” or “habitat”. Each area is characterized by
Finite impulse response (FIR) filter are the digital filter some features which include position, availability of
which always produce a stable output. FIR filter’s response vegetation, rainfall etc, and are known as habitat suitability
can be given as given in (1). index (HSI). The number of species which a habitat hosts also
helps in determining the value of HSI because it is in direct
H ( z )  h(0)  h(1) z 1  ...  h( N ) z  N (1) proportion to the HSI’s value. The habitability is defined by
N
certain variables referred as suitability index variables (SIVs).
or, H ( z )   h( n) z
n 0
N
(2) These can be termed as independent variables while HSI are to
be termed as dependent variables. The habitats having high
value of HSI resist changes but are ready to dispense their
where h(n) is the impulse response and N is order of the features with the habitat with low HSI. Similarly habitats with
filter. Filter’s length can be determined from N+1. The low HSI are more vulnerable to changes and are ready to
designing process is chosen so as to obtain impulse response accept features from habitats with high HSI. The entering of
of filter. species into the habitat is termed as immigration and the
exiting of species from the island is termed as emigration.
Filter’s frequency response can be formulated as,
N
The algorithm works with two operators namely Migration
H (e jwk
)   h( n)e  jwk n
(3)
operator and Mutation operator. Migration operator adjusts the
habitat by moving the SIVs of better habitat into it which are
n 0
chosen according to their immigration rate (λ) and emigration

where  = , H( 
) is the Fourier transform vector. rate (μ) thus improving the habitat and further improving the
 generation. Mutation operator modifies the SIVs according to
The frequency in [0, π] is sampled with N points. The error
probability which is based on λ and μ parameters. Fig. 1
function refers to the difference between the magnitude of an
shows the variation of emigration rate with immigration rate
ideal filter and the designed filter. Thus, it can be formulated
with the number of species in single habitat. ‘I’ is the
as,
maximum immigration rate whereas ‘E’ is the maximum
N emigration rate, which are usually taken as one. ! is the
Error  max{[|| H i (e jwi ) |  | H d (e jwi ) ||]} (4) maximum number of species a habitat can handle making the
n 0 immigration rate minimum i.e. 0 and emigration rate to
maximum i.e. 1.
E ()  G()[ Hi (e jwi )  H d (e jwi )] (5)

where G(ω) is the function which provide weights for


different frequency bands and is termed as weighting function;
(  ) is the frequency response of ideal filter and it can be
defined as,
1 ,1 ≤ ω ≤ 
H i (e jw )   0 , ℎ (6)

where  is cut-off frequency of the designed filter and


 (  ) is the frequency response of designed filter. The
error function needs to be minimized so fitness function has
been reformulated as
J  max p (| E () |  p )  max p (| E() |  s ) (7) Fig. 1. Relationship of immigration and emigration rates with number of
species in one habitat
where  and  are the pass band and the stop band ripple The values of λ and μ can be calculated as,
values;  and  are the normalized edge frequencies in the
EK
pass band and stop band respectively.
 (8)
The error function in (7) gives a generalized error function
P
which has to be minimized by proposed BBO. This function K
will be used to optimize the filter’s performance.    (9)
P
where E=maximum value of λ; I= maximum value of μ;
P=population size; K=number of species of the K-th

313
individual. The pseudo-code of the migration and mutation the habitat [15]. The migration operator for this
operators is given in following part. modified version can be given with the following
pseudo-code.
Algorithm1:Pseudo-code for migration operator
1. For r =1 to P Algorithm3:Pseudo-code for BBO with chaos

2. Choose " with probability proportional to λ" 1. For r =1 to P

3. If " is chosen 2. Choose " with probability proportional to $"

4. For s=1 to n 3. If rand < $"

5. Choose  with probability proportional to μ" 4. For s =1 to n

6. If  is selected 5. Choose H with probability proportional to μ"

7. Choose a random SIV σ from  6. If rand < μ"

8. Substitute a random SIV in " with σ 7. Choose random SIV σ from 

9. End if 8. " =   (&.'('*+- ))

10. End for 9. End if

11. End if 10. End for

12. End for 11. End if

Algorithm2:Pseudo-code for mutation operator 12. End for


1. For r=1 to P  Enhanced BBO: Standard migration operator creates
similar habitats which lead to duplication of habitats.
2. For s=1 to n
The duplication ultimately reduces the diversity of the
3. Use $" and μ" and calculate %" habitat. To avoid this harmful redundancy, a modified
operator termed as ‘modified clear duplicate operator’
4. Choose SIV " (s) with probability similar to %" is introduced to clear the duplicates into basic code
5. If " (s) is selected which increases the computation time of the algorithm
but with a considerable increase in its performance.
6. Substitute " (s) with randomly generated SIV This operator clears the duplicate habitats by selecting
7. End if a value between minimum and maximum value of
SIVs [16]. The proposed migration operator can be
8. End for coded in the pseudo-code as follows.
9. End for Algorithm4:Pseudo-code for enhanced BBO
Here P=size of population; n=number of variables in a 1. For r = 1 to P
habitat; $" =immigration rate; μ" =emigration rate ; "
=emigrating habitat;  =immigrating habitat; 2. For s = s+1 to P
%" =immigration probability; σ=immigrating SIV. 3. If " = 
A. BBO Variants 4. " = (min ( " ) + (max (  ) - min ( " ))*rand)
BBO’s migration operator exploits the habitats by fine- 5. End if
tuning the SIVs of that habitat. This means previous
solutions are updated while mutation operator is the 6. End for
operator which generates new SIVs. In this paper, the 7. End for
variants have been modified in the migration operator of
the algorithm because the major work in BBO is Here rand= any random integer between 0 and 1
performed by the migration operator. So, mutation  Blended BBO: In this variant of BBO, new solutions
operator is not modified here to check the performance of are made up of two parts: feature from other solution,
migration operator variants. The variants analyzed in this and feature from itself. This technique has been
paper are given below with their modified migration adopted from blended crossover from GAs. Thus the
operator. upcoming generation consists of own parent’s feature
 BBO with Chaos: In this variant of BBO, real coding as well as migrating SIV’s feature. A multiple ‘α’ has
is adopted. The migration operator is adopted from been introduced whose value is between 0 and 1 or it
strength parent evolutionary algorithm (SPEA). The could be any random or deterministic value [14]. The
individuals possessing high HSI share their pseudo-code for this BBO variant can be presented as
characteristics with chaos and improve the diversity of given in following part.

314
Algorithm5:Pseudo-code for blended BBO The values of various parameters used for construction of
filters are:
1. For r =1 to P
 passband frequency ( ) = 0.75;
2. Choose " with probability proportional to λ"
 stopband frequency ( ) = 0.65;
3. If " is chosen
 passband ripples ( ) =0.1;
4. For s=1 to n
5. Choose  with probability proportional to μ"  stop band ripples ( ) = 0.01;

6. If  is chosen  Population size in BBO = 40;

7. " = α ( " ) + (1- α)   Maximum number of generations = 200;

8. End if  Dimension = 11;


9. End for  Probability of mutation = 0.01;
10. End if  Elitism parameter=2.
11. End for The code has been evaluated and run on MATLAB 2012b
version on core (TM) i3 processor, 3.00 GHz with 3 GB
Here α=0.5. RAM. Fig. 2 shows the magnitude response of all the four
 BBO with immigration refusal (IR): Migration variants i.e. BBO with chaos, enhanced BBO, blended BBO
operator improves the SIVs by immigrating high HSI and BBO with IR over same objective function of high pass
habitats to low HSI habitat. But there are chances that filter. The frequency parameter has been normalized in the
there is immigration from low-fitness solution to high figure. Table I shows the optimized coefficients for the
fitness solution thus ruining the fitness of island. To designed FIR filter having order 20. Meanwhile, Table II
tackle this problem, fitness of immigrating island is presents the best, worst and average fitness value of all the
taken into consideration and is rejected if it is lower modified variants. The values from table and the graph
than the fitness of emigrating island [18]. The code for suggest that best frequency response has been produced by the
this variant is as follows: modified variant named blended BBO. The frequency
response graph suggests that other variants have same
Algorithm6:Pseudo-code for BBO with IR response in transition band but there is an improved response
1. For r= 1 to P in stop band region by blended BBO. This is because it makes
use of two features: the parents and the immigrating habitat’s
2. Choose " with probability proportional to λ" features which results in least fitness value and least ripples as
seen in Fig. 2. BBO with chaos also uses both features same as
3. If " is chosen
blended BBO but with exponent which makes the difference
4. For s=1 to n in computation thus making blended BBO the best variant of
BBO in optimization of FIR high pass filter.
5. Choose  with probability proportional to μ"
B. Comparison on basis of convergence curves
6. If fitness ( " ) < fitness (  )
To compare each modified variant on the basis of
7. Substitute " with  convergence profile, convergence curves have been drawn in
8. End if Fig. 3 by plotting the average fitness values for a given
number of iterations. The iteration number is selected as 200
9. End for for each modified variant.
10. End if Fig. 3 concludes that blended BBO converges at a much
faster pace as compared to other variants of BBO. Also, it
11. End for
attains a much lower value of fitness function than other
IV. RESULTS AND COMPARISONS modified variants in less number of iterations. Meanwhile,
enhanced BBO produces a sub-optimal value for the fitness
A. Study of magnitude response of FIR highpass filter function and converges at a slower pace than blended BBO.
In order to bring out the results and compare the The reason for this is that enhanced BBO remove the habitats
effectiveness of all modified variants, FIR filter has been which result in reduction of diversity. BBO with IR attained a
designed using MATLAB. The designed filter has an order of low value which may be due to the refusal of SIVs.
value 20 thus the number of coefficients is 21. The number of Furthermore, BBO with chaos did not come up with good
sampling points is taken as 256 and the sampling frequency is results which might be due to the intake of parent SIVs but in
chosen as 1 Hz. Each modified variant of BBO has been run the exponential form as given in Algorithm 3. Thus, blended
40 times to get optimum solution. BBO brought out better results as compared to other variants
of BBO i.e. enhanced BBO, BBO with immigration refusal
and BBO with chaos.

315
TABLE I. OPTIMUM COEFFICIENTS OF BBO VARIANTS
h(N) BBO with chaos Enhanced BBO Blended BBO BBO with IR
h(1)=h(21) - 0.0172191591766668 0.0000000000000000 0.000808732857412956 -0.00135941944083007
h(2)=h(20) 0.0000000000000000 0.0200000000000000 0.0188318492348605 0.0000000000000000
h(3)=h(19) 0.0181391428940318 -0.0442181142862795 -0.0407884508250713 0.0368361088292808
h(4)=h(18) 0.0000000000000000 0.00403604349295295 0.00852423732438976 -0.0138256096358770
h(5)=h(17) 0.0000000000000000 0.0244260303745714 0.0380592939106986 0.0000000000000000
h(6)=h(16) 0.0531976107751360 -0.0521174300633237 -0.0592740463931042 0.0450974662849897
h(7)=h(15) -0.0768191037339498 0.0544200854784211 0.0464247337218026 -0.0867299707114245
h(8)=h(14) 0.0000000000000000 0.0300000000000000 0.0325544425941253 0.000000000000000
h(9)=h(13) 0.207746275696171 -0.160498653596845 -0.134532387021371 0.191909339460792
h(10)=h(12) -0.275936628556418 0.247392588790788 0.270756153116677 -0.237300806497940
h(11) 0.286528896803798 -0.307971789732571 -0.278649888479125 0.313559044851423

TABLE II. COMPARISON OF BEST WORST AND MEAN VALUES

BBO Variant Best Value Worst Value Mean Value


BBO with 0.2498 2.6187 0.6551
chaos
Enhanced 0.0621 2.6796 0.2456
BBO
Blended BBO 0.0616 2.5451 0.2175
BBO with IR 0.1877 2.4374 0.4633

Fig. 2. Magnitude and frequency graph for different variants of BBO.

Fig. 3. Convergence profile of different variants of BBO.

316
V. CONCLUSION [14] H. Ma and D. Simon, “Blended biogeography-based
optimization for constrained optimization,” Engineering
In this paper, FIR high pass filter has been Applications of Artificial Intelligence, vol. 24 , pp. 517-
optimized with various modified variants namely, 525, 2011.
BBO with chaos; enhanced BBO; blended BBO; [15] X. Wang and Z. Xu, “Multi-Objective Algorithm based on
BBO with IR. The simulation results showed that Biogeography with Chaos,” International Journal of
out of all the variants, blended BBO gives out Hybrid Information Technology, vol. 7 , pp. 225-234,
2014.
better results. The convergence profile also proved
[16] S. Singh and G. Sachdeva, “Yagi-Uda Antenna Design
blended BBO as a better option for the proposed Optimization for Maximum Gain using different BBO
objective. For future work, modifications can also Migration Variants,” International Journal of Computer
be brought out in the mutation operator of BBO Applications, vol. 58 , pp. 86-96, 2012.
and these variants can be applied to problems other [17] S. Singh, Shivangna and S. Tayal, “Analysis of Different
than filters. Ranges for Wireless Sensor Node Localization using PSO
and BBO and its variants,” International Journal of
Computer Applications, vol. 63, pp. 31-37, 2013.
[18] U. Singh, H. Kumar and T.S. Kamal, “Design of Yagi-Uda
VI. REFERENCES Antenna Using Biogeography Based Optimization,” IEEE
Transactions on Antennas and Propagation, vol. 58, pp.
[1] L. Litwin, “FIR and IIR digital filters,” IEEE Potentials,
3375-3379, 2010.
vol. 19, pp. 28-31, 2000.
[19] U. Singh, H. Kumar and T.S. Kamal, “Linear Array
[2] N.E. Mastorakis, I.F. Gonos and M.N.S. Swamy, “ Design
Synthesis using Biogeography Based Optimization,”
of two-dimensional recursive filters using genetic
Progress In Electromagnetics Research M, vol. 11, pp. 25-
algorithms,” IEEE transaction on Circuits and systems-I:
36, 2010.
Fundamental Theory and Applications, vol. 50, pp. 634-
639, 2003. [20] U. Singh and T.S. Kamal, “Design of non-uniform circular
antenna arrays using biogeography based optimization,”
[3] N. Karaboga and B. Cetinkaya, “Design of FIR filters
Microwaves, Antennas & Propagation, IET, vol. 5, pp.
using differential evolution algorithm,” Circuits, Systems
1365-1370, 2011.
and Signal Processing, vol. 25, pp. 649-660, 2006.
[21] U. Singh and T.S. Kamal, “Optimal synthesis of thinned
[4] J.I. Ababneh and M.H. Bataineh, “Linear phase FIR filter
design using particle swarm optimization and genetic array using biogeography based optimization,” Progress In
Electromagnetics Research M, vol. 24, pp. 141-155, 2012.
algorithms,” Digital Signal Processing, vol. 18 , pp. 657-
668, 2008. [22] U. Singh and T.S. Kamal, “Synthesis of thinned planar
[5] B. Luitel and G.K. Venayagamoorthy, “Differential circular array antennas using biogeography based
evolution particle swarm optimization for digital filter optimization”, Emerging Technology Trends in
design,” , IEEE World Congress on Computational Electronics, Communication and Networking
Intelligence, pp. 3954-3961, 2008. (ET2ECN),2012, IEEE, pp. 1-5, 2012.
[6] D. Simon, “Biogeography-based optimization,” IEEE [23] U. Singh and T.S. Kamal, “Concentric Circular Antenna
Transactions on Evolutionary Computation, vol. 12, pp. Array Synthesis using Biogeography Based Optimization,”
702-713, 2008. Majlesi Journal of Electrical Engineering, vol. 6, pp. 48-
55, 2012.
[7] G. Wang, L. Guo, H. Duan, L. Liu and H. Wang,
[24] U. Singh, “Design of Concentric Circular Antenna Array
“Dynamic Deployment of Wireless Sensor Networks by
using Biogeography Based Optimization,” International
Biogeography Based Optimization Algorithm,” Journal of
Sensor and Actuator Networks, vol. 1, pp. 86-96, 2012. Journal of Enhanced Research in Science Technology &
Engineering, vol. 2, pp. 119-125, 2013.
[8] D. Du, D. Simon and M. Ergezer, “Biogeography-Based
Optimization Combined with Evolutionary Strategy and
Immigration Refusal,” Systems, Man and Cybernatics,
pp.997-1002 , 2009.
[9] H. Ma and D.J. Simon, “Biogeography-Based
Optimization with Blended Migration for Constrained
Optimization Problems,” Genetic and Evolutionary
Computation Conference, pp. 417-418, 2010.
[10] S.S. Pattnaik, M.R. Lohokare and S. Devi, “Enhanced
Biogeography-Based Optimization using Modified Clear
Duplicate Operator,” Nature and Biologically Inspired
Computing, pp. 715-720, 2010.
[11] S. Mondal, Vasundhara, R. Kar, D. Mandal and S.P.
Ghoshal, “Linear Phase High Pass FIR Filter Design using
Improved Particle Swarm Optimization,” World Academy
of Science, Engineering and Technology, vol. 60, pp.
11682-11689, 2011.
[12] S. Mandal, S.P. Ghoshal, P. Mukherjee, D. Sengupta, R.
Kar and D. Mandal, “Design of Optimal Linear Phase FIR
High Pass Filter using Improved Particle Swarm
Optimization,” ACEEE International Journal on Signal
and Image Processing, vol. 3 , pp. 5-9, 2012.
[13] S. Mandal, S.P. Ghoshal, R. Kar and D. Mandal, “Design
of optimal linear phase FIR high pass filter using craziness
based particle swarm optimization technique,” Journal of
King Saud University- Computer and Information
Sciences, vol. 24 , pp. 83-92, 2012.

317