TELKOMNIKA Telecommunication Computing Electronics and Control

Vol. 20, No. 5, October 2022, pp. 955~962

ISSN: 1693-6930, DOI: 10.12928/TELKOMNIKA.v20i5.24090  955

Clutter reduction technique based on clutter model for

automatic target classification in forward scatter radar

Nur Emileen Abd Rashid1, Nor Ayu Zalina Zakaria2, Zuhani Ismail Khan1,
Siti Amalina Enche Ab Rahim2, Nur Luqman Saleh3
1
Microwave Research Institute, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia
2
School of Electrical Engineering, College of Engineering, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia
3
Wireless and photonic network (WiPNET), Universiti Putra Malaysia, Serdang, Seri Kembangan, Selangor, Malaysia

Article Info ABSTRACT

Article history: Classification becomes one of the important elements in the forward scatter
radar (FSR) micro-sensors network. This classification performance is
Received Aug 24, 2021 dependent on the target’s profile behaviour and the network’s surrounding;
Revised Aug 09, 2022 and one of the factors that cause the reduction of classification probability is
Accepted Aug 20, 2022 the presence of ground clutter. As the volume of clutter increases, their
masking effect becomes greater and may result in more significant errors in
target classification. Hence, to reduce misclassification in the FSR sensor
Keywords: network, a new clutter reduction technique based on the ground clutter
model is proposed. Simulated ground clutter is modeled based on the
Clutter reduction technique estimated signal to clutter ratio (SCR) of the received signal. The clutter
Forward scatter radar effect is diminished by eliminating simulated like-clutter from the receiving
Radar clutter signals. The result shows improvement in the classification accuracy,
Target classification especially for the minimum value of the SCR and this new technique uses
only one database which will shorten the processing time and reduce the
overall database’s size.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Nur Emileen Abd Rashid
Microwave Research Institute, Universiti Teknologi MARA
40450 Shah Alam, Selangor, Malaysia
Email: [email protected]

1. INTRODUCTION
The performance of forward scatter radar (FSR) for detecting and classifying ground targets (human
and vehicles) has been a subject of investigation in recent years [1]-[12]. When a target crosses the radar
baseline, the radio wave is partially obstructed and disturbed, resulting in a Doppler signature at the receiver.
This signal can be used to detect targets, estimate parameters, track them, and perhaps perform automatic
target classification (ATC).
The FSR micro-sensors system is capable of correctly detecting and classifying ground targets,
particularly under ideal or controlled settings [4]-[8]. Numerous algorithms for pattern recognition have been
developed, including the K-nearest neighbour (KNN) and neural networks (NN) classifiers. In [6]
demonstrated that by combining the output of principal component analysis (PCA) with the input of a KNN,
the FSR classification performance above 90% accuracy even at low frequency. Following that, [10], [12]
discussed how to improve classification performance using NN. The network’s inputs are the Z-score
parameters such as mean and standard deviation. While in [11], the time domain signal is analysed using the
wavelet transformation approach for the goal of identification and classification.
The aforementioned techniques provide a strong indication of the classification system’s
performance. Nonetheless, the classification system’s performance is influenced by the target’s profile

Journal homepage: http://telkomnika.uad.ac.id

956  ISSN: 1693-6930

behaviour, which includes factors such as speed, crossing angles, crossing points, baseline length, and sensor
placements, as well as the presence or absence of clutter [13], multiple targets [14] and other factors. This
article will only discuss the effect of clutter on the FSR system and how it may be used to improve
classification performance.
Clutter is an undesired signal from the environment that originates either internally or externally.
Ground clutter may be caused by vegetation, precipitation such as rain, snow, or hail, as well as interference
and multipath caused by buildings, the ground, and the sea. The FSR micro-sensors network distributes
sensors on the ground with a baseline length of hundreds of metres and also makes use of omnidirectional
antennas at extremely low elevations. The existence of swinging foliage and wind-blown branches nearby
sensors could be considered volume-distributed scatterers whenever the sensors are surrounded by
vegetation. Due to the non-directional nature of the antennas and the lack of range resolution, clutter
accumulated from a bigger volume.
The impact of clutter or noise on the FSR classification systems have been studied previously [15], [16].
Numerous signal-to-clutter ratios (SCR) have been simulated from 0 dB to 30 dB where few possible sources
of error were taken into accounts such as erroneously estimate the speed of the target and target spectrum’s
shape deformation. Changes in the classification performance can be perceived as SCR level is decreased;
which could cause false classification. It has shown almost no effect on the classification accuracy (except
for the lowest SCR) due to speed estimation error, which exhibits the robustness of the speed estimation
algorithm towards clutter. However, when it comes to the spectrum’s shape deformation error, the effect can be
detected even when the clutter power is small. Based on this analysis, it can be concluded that the main source
of error which can degrade ATC performance is due to the target’s spectrum shape deformation. By taking into
account both sources of error at the same time, it was discovered that the classification performance declines
dramatically for SCR below 20 dB, particularly at 64 MHz, when the classification accuracy plummeted from
80% to almost 20%. Consequently, the clutter compensation technique has been introduced to eliminate this
effect. Different database for a different level of clutter has been implemented. This new technique demonstrates
some enhancements where the classification accuracy increases by 20% for SCR = 0 dB. However, this
technique introduces a large size of the database which contributes towards longer processing time and a
lower percentage of accuracy (less than 50%) for SCR below 15 dB.
Hence, in this paper, a new technique is proposed which provides a promising improvement in the
system performance while considering a more realistic solution by implementing the clutter reduction
technique (CRT) based on the ground FSR clutter model. By using this technique, only one database which
consists of original target signatures is used. The paper starts with describing FSR configuration and ATC,
followed by a discussion on the ground FSR clutter model. Then, the improvement of the ATC system is
presented and discussed including the integration of the ground FSR clutter model with clutter reduction
technique. The final section summarizes the outline of future work in the FSR network research.

2. RESEARCH METHOD
2.1. Measurement setup and signal processing
On an empty concrete field with a baseline distance of 50 m, an outdoor experiment was conducted
to gather target signals in a clutter-free environment (assumed to have a low vegetation clutter level).
Four continuous wave carrier frequencies are simultaneously transmitted in the very high frequency (VHF)
(60 MHz – 150 MHz) and ultra high frequency (UHF) (400 MHz) bands. This experiment incorporated 200
signals from four different vehicle types: vehicle 1 (4.3 m × 1.4 m, hatchback), vehicle 2 (4.4 m × 1.4 m,
hatchback), and vehicle 3 (4.7 m × 1.9 m, sport utility vehicles (SUV)). In this research, the system’s ability
will be evaluated to distinguish between similar vehicles, particularly when certain characteristics are present
especially when specific applications are applied for example road statistic or automatic tolling system.
The outdoor measurement is conducted in the manner depicted in Figure 1. The testing scenario
begins when a target moves with a velocity described by projections 𝑣𝑥 and 𝑣𝑦 , and crosses the baseline at
the position (0, 𝑦𝑐 ) with a motion direction provided by angle. As the target approaches the baseline, the
range between the transmitter-target (𝑅𝑇 ) and target-receiver (𝑅𝑅 ) fluctuates with time, resulting in a phase
(Doppler) shift in the received signal. The received signal can be presented as:

𝑆𝑟 (𝑡) = 𝐴(𝑡) 𝑠𝑖𝑛[ 𝜓(𝑡)] (1)

The target signature, 𝐴(𝑡) varies with target radar cross section (RCS),  which is entirely specified by
the target geometrical cross-section (silhouette), propagation loss, 𝐿 𝑇 (𝑡), 𝐿𝑅 (𝑡) along the transmitter-target and
target-receiver paths. Other parameters involved in target detection are the transmit power, 𝑃𝑡 and the antenna
gains, 𝐺𝑇 and 𝐺𝑅 where it remains constant. The envelope of the received signal is written as:

TELKOMNIKA Telecommun Comput El Control, Vol. 20, No. 5, October 2022: 955-962
TELKOMNIKA Telecommun Comput El Control  957

4𝜋𝜎(𝑡)
𝐴(𝑡) = √𝑃𝑇 𝐺𝑇 𝐺𝑅 𝐿 𝑇 (𝑡)𝐿𝑅 (𝑡) (2)
𝜆2

By utilising signal processing techniques, it is possible to extract important information for

classification, such as the target’s speed and feature. When a target crosses the transmitter-receiver baseline,
the sensors can relay the pre-detection signal to headquarters over a wireless network. The received signal
will be processed further in the ATC algorithm block, where it will contain information on the target’s
silhouette and speed.
The signal that has been received passes through a process called the fourier transform (FT), which
converts it into the frequency domain. In this domain, the spectrum displays the Doppler components that are
present in the signal. Technically speaking, the speed of the target has an effect on the spectrum since the
major lobe of the spectrum’s bandwidth depends on the speed 𝑣, as well as the length of the target 𝑙, and the
equation for this is:
𝑣
∆𝑓 = (3)
𝑙

As a result, pre-processing of the spectrum is required before it can be used as a feature vector in the
classification system. The target is then classified using a feature vector. The spectral feature vector obtained
at the speed normalisation block’s output has a high degree of dimension and the features are highly linked.
As a result, principal component analysis (PCA) is utilised to reduce the dimensionality of the data. PCA’s
output is then used for classification, either as training or testing data.
In general, a classification system’s standard approach is divided into two stages: training and
testing. During the training stage, a model for each vehicle is constructed using the training data set (this is
done using the power spectrum density (PSD) values of the entire feature vector), and during the testing
stage, a PCA-based feature vector is extracted from an unknown received vehicle signal. However, due to the
scarcity of data in this research, all recorded signals were employed alternately in classification trials rather
than dividing the database into training and testing sections. To make it feasible, a leave-one-out approach
was adopted, in which only one vehicle signal was used for testing at a time, while the remaining signals
were used for training.
After that, the Euclidean distance between this feature vector and all feature vectors associated with
trained models is determined. The test data is sorted into the most prevalent class among its K-nearest
neighbours. If the allocated group and the test data’s actual group match, the classification is termed
successful.

Figure 1. The outdoor measurement topology

2.2. Errors due to the presence of clutters in FSR system

A time-domain simulated non-stationary clutter signal is added to the received vehicle signature at
different SCRs (from 0 dB to 30 dB, in 5 dB steps) in order to introduce the effect of clutter into the FSR system
and ensure continuous improvement. Low SCR values are used in this research to simulate the worst-case
scenario, particularly in the presence of natural disasters, where the clutter signal is greater than the target’s
Doppler. The SCR is defined as the (4).
𝑃
𝑆𝐶𝑅 = 10 log( 𝑠 ) (4)
𝑃𝑐

Where 𝑃𝑠 and 𝑃𝑐 are the total power of the signal and total power of the clutter, respectively.

Clutter reduction technique based on clutter model for … (Nur Emileen Abd Rashid)
958  ISSN: 1693-6930

The clutter received signal, 𝑆𝑐 (𝑡) is a randomly generated clutter received signal which is generated
by employing the clutter simulation model described in [17]-[19]. The signal is then bypassing a white
Gaussian noise through a low-pass filter with a cut-off frequency that varies depending on the frequency
being used; this generates simulated clutter which is similar to the measured clutter spectrum. The simulated
cluttered received signal, 𝑆𝑐𝑠 (𝑡) is given by (5).

𝑆𝑐𝑠 (𝑡) = 𝑆𝑟 (𝑡) + 𝑆𝑐1 (𝑡) (5)

Where 𝑆𝑟 (𝑡) is the received signal from target detection in section 2.1, (1).
As mentioned earlier, the presence of clutter in the target’s signal introduces two possible sources of
error in the ATC system: error due to speed estimation and error due to signature deformation. The speed
estimation error is the inaccuracy caused by the speed estimation algorithm estimating the speed incorrectly.
In the real world, a cluttered received signal is used to determine the target’s speed and as an input to the
classification process. As previously stated, the bandwidth of the spectrum’s main lobe 𝛥𝑓 is dependent on
the speed 𝑣 and the length of the target l; consequently, when the speed normalisation method is applied, the
spectrum may not be appropriately rescaled due to the erroneous ratio between the estimated and reference
speeds. Clutter level diverges with the surroundings. As the clutter level varies, the amplitude of the signal
fluctuates and the shape of the signal distorts which introduces signature deformation error. As the SCR level
lowers, the PCA components for each type of target become more spread out, resulting in increased overlap across
the various targets. This leads to a decrease in classification performance. The reader is prompted at [18], [20]
for more technical information on the subject matters.

2.3. CRT based on ground fsr clutter model

Any classification system’s performance rapidly declines as the discrepancy between training and
test conditions increases. Numerous efforts have been made to correct for the effect of noise in a variety of
domains of pattern recognition, including automatic voice recognition. Numerous methods for minimizing
the influence of noise have been proposed, including filtering prior to classification [20], [21], model-based
noise compensation [22], [23], noise-resistant feature extraction [24], and numerical simulation [25].
This article proposed a novel technique for compensating for clutter that is based on the FSR ground
clutter model. After receiving the cluttered received signal, it is necessary to acquire an estimate of the noisy
conditions. The ratio of signal to clutter energy, SCR, is determined using (4). As illustrated in Figure 2,
the energy of the clutter is then evaluated using signal samples taken before and after the target signal is
identified.
Another set of random clutter is simulated using the clutter modelling parameters [17], [19].
The clutter power level 𝑃𝑠𝑐2 is modified in accordance with the computed SCR and signal power of the
received clutter signal acquired in the preceding step. The simulated clutter signal 𝑆𝑐2 (𝑡) is then subtracted
from the cluttered received signal, where the subtracted clutter signal 𝑆𝑠𝑐 (𝑡) is defined as:

𝑆𝑠𝑐 (𝑡) = 𝑆𝑐𝑠 (𝑡) − 𝑆𝑐2 (𝑡) (6)

𝑆𝑠𝑐 (𝑡) is then run through a bandpass filter to remove any leftover noise and smooth the signal
before continuing with the classification procedure. The denoised signal, 𝑆𝐷 , is the result of this procedure.
Figure 2 illustrates the time-domain signals for the original, cluttered, and denoised signals.

Figure 2. Target signals: original, with 0 dB clutter and after denoising process

TELKOMNIKA Telecommun Comput El Control, Vol. 20, No. 5, October 2022: 955-962
TELKOMNIKA Telecommun Comput El Control  959

Simultaneously, the target’s speed is measured and the denoised data is converted to the power
spectrum. Prior to further processing, the feature that contains enough information to discriminate between
distinct classes and is as noise-resistant as possible is chosen. The frequency range of the spectrum is
determined using knowledge about the signal’s and clutter’s qualities. It is well established that clutter is
concentrated in the low frequency region of the spectrum between 0.5 Hz and 1 Hz of the Doppler band, but
the target signature may occupy a broad Doppler range.
Additionally, to the preceding approach, there is no requirement to assess the complete spectrum of the
target for ATC purposes. This is because the difference between different sorts of targets is visible only at the
power spectrum’s highest point, which is within 20 dB of its maximum value of 0 dB; roughly 0.5 Hz − 7 Hz
(depends on the carrier frequency). As a result of this information, the extremely low and very high frequency
bands of the spectrum are omitted from the subsequent stages of feature extraction. As a result, the
classification system receives as input only the highlighted portion of the signal spectrum.

3. RESULTS AND ANALYSIS

Four different vehicle signals at 3 different frequencies are utilised to determine the efficacy of the
clutter reduction strategy based on the clutter model. Previously, the clutter-reduction technique is based on
several SCR databases is applied. In this study, a single database with SCR = 30 dB (original measured
signal) is used, but other parameters such as the number of k for KNN and the number of PCAs (𝑛𝑃𝐶𝐴) stay
constant (𝑘 = 3 and 𝑛𝑃𝐶𝐴 = 3).
Plots of data from various SCR levels are made against this database. PCA plots for SCR = 0 dB and
SCR = 30 dB are shown in Figure 3(a) and Figure 3(b). The figures demonstrate that, with the exception of a
few outliers, the positions of both SCRs data are highly correlated. This discrepancy could be explained by
residual clutter in the signal, particularly at SCR = 0 dB.

(a) (b)

Figure 3. PCA plots for a different level of SCRs after clutter reduction technique at frequency 151 MHz when
(a) SCR 0 dB and (b) SCR 30 dB, (notation: vehicle 1 ( ), vehicle 2 (◊), vehicle 3 (O), and vehicle 4 (+))

Table 1(a), Table 1(b), and Table 1(c) compare the percentage of correctly classified instances for
four different types of vehicles at three different frequencies and with varying SCR levels. The tabulated
statistics demonstrate the classifications’ exceptional accuracy, particularly for vehicle 4, where 100%
classification was achieved at 151 MHz. This is clearly related to vehicle 4’s physical shape, which is larger
and has more defined edges (rectangular shape) than other cars. Apart from that, for SCR values more than
20 dB, the results indicate a high degree of classification accuracy, except for vehicle 1 at 64 MHz and
vehicle 3 at 434 MHz. This is because the design and dimensions of both vehicles are close to those of
vehicle 2, and hence both were incorrectly classed as vehicle 2. While 434 MHz should theoretically provide
a larger radar cross-section and more frequency information about the target signal, and 64 MHz should
provide lesser propagation loss and clutter, what is fascinating about this data is that 151 MHz performs the
best of these frequencies. This could be because 151 MHz has less propagation loss/clutter than 434 MHz
and more frequency information than 64 MHz. Thus, 151 MHz is the optimal frequency because it
compromises betweem 64 MHz and 434 MHz.

Clutter reduction technique based on clutter model for … (Nur Emileen Abd Rashid)
960  ISSN: 1693-6930

Table 1. Vehicle classification confusion matrices at (a) 64 MHz, (b) 151 MHz, and (c) 434 MHz
Vehicle classification confusion matrices SCR Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4
(a) 64 MHz 0 55 98 69 84
5 75 98 86 84
10 80 98 86 96
15 83 98 88 96
20 83 98 88 96
25 83 98 90 96
30 83 98 90 96
(b) 151 MHz 0 60 82 63 68
5 85 90 63 88
10 83 92 82 92
15 88 92 84 96
20 93 92 88 96
25 93 92 88 96
30 93 92 88 96
(c) 434 MHz 0 73 96 90 92
5 88 98 92 98
10 88 98 94 100
15 93 98 94 100
20 93 98 94 100
25 93 98 94 100
30 93 98 94 100

Figure 4 depicts the average percentage of classification accuracy using automatic target
classification-clutter compensation (ATC-CC), with ATC-CC and CRT at different frequencies. As can be seen
from the graph, classification performance improves as the SCR increases. Interestingly, the clutter reduction
strategy significantly improves the classification accuracy of 64 MHz, particularly for SCR = 0 dB − 5 dB,
where accuracy increases by more than 50% and continues to improve as SCR grows. As the graph indicates,
the same performance is observed at 151 MHz and 434 MHz. However, for SCR values more than 20 dB at
151 MHz, when the clutter impact is negligible, the result indicated that improved performance can be
observed without using CRT. This is due to the fact that the database was not corrupted by simulated clutter.
Additionally, as compared to other frequencies, 151 MHz strikes a balance between clutter and data gathered
from the target signature. Overall, these results indicate that an increase in classification accuracy of at least
30% is possible for signal-to-noise ratio (SNR) less than 10 dB at all frequencies.
Additionally, the CRT based on the FSR ground clutter model outperforms the clutter compensation
(ATC-CC) model. As demonstrated in the graph, with SCR = 0 dB, a greater than 30% improvement in
accuracy can be observed at all frequencies for CRTs with a percentage of accuracy greater than 68%. CRT
performance improves as the number of SCR grows, reaching 90% accuracy for SCR greater than 15 dB.
Interestingly, both approaches operate comparably well at 151 MHz, as the SCR is greater than 20 dB.
However, at other frequencies, the FSR ground clutter model CRT outperforms ATC-CC. Overall, our results
demonstrate that CRT significantly improves classification of ground vehicles, even those with similar
shapes and dimensions and operating at varying frequencies, without requiring a large database.

Figure 4. Comparison of classification accuracy before and after the use of ATC-CC, as well as between
ATC-CC and CRT

TELKOMNIKA Telecommun Comput El Control, Vol. 20, No. 5, October 2022: 955-962
TELKOMNIKA Telecommun Comput El Control  961

4. CONCLUSION
This paper presents a novel technique for decluttering based on the clutter model. The results
demonstrate the technique’s efficacy in a cluttered FSR environment. Even when the power clutter is
identical to the strength of the vehicle’s signal with SCR equal to 0, the minimum classification accuracy is
around 70%. Clutter reduction techniques based on the FSR clutter model can help reduce overall processing
time and database size. Additionally, regardless of the amount of clutter in the signal, the system can be
automatically adjusted to the present condition, which minimises the system’s complexity indirectly. For future
work, other clutter reduction approaches or classification strategies can be investigated to increase the ATC
system’s performance when there is a high level of clutter in the surrounding.

ACKNOWLEDGEMENTS
This research is fully supported by UiTM grant, 600-RMC/MYRA 5/3/LESTARI (009/2020). The
authors would like to thank Research Management Institute (RMI) and Universiti Teknologi MARA for all
the supports.

REFERENCES
[1] R. S. A. R. Abdullah, M. F. A. Rasid, M. W. Azis, and M. Khalafalla, “Target prediction in Forward Scattering Radar,” 2007
Asia-Pacific Conference on Applied Electromagnetics, 2007, pp. 1-5, doi: 10.1109/APACE.2007.4603914.
[2] C. Hu, T. Long, C. Mikhail, and X. L. Li, “Forward scattering micro radars for situation awareness,” 2009 2nd Asian-Pacific
Conference on Synthetic Aperture Radar, 2009, pp. 231-234, doi: 10.1109/APSAR.2009.5374099.
[3] A. D. Luca, “Forward Scatter Radar: Innovative Configurations and Studies,” PhD Thesis, University of Birmingham, 2018.
[Online]. Available: https://etheses.bham.ac.uk//id/eprint/8669/1/De_Luca18PhD.pdf
[4] C. Zhang, C. Liu, C. Hu, and X. Liu, “Statistical analysis and simulation method of forward scattering clutter in bistatic radar,”
IET International Radar Conference 2013, 2013, pp. 1-6, doi: 10.1049/cp.2013.0479.
[5] C. Hu, T. Zeng, and C. Zhou, “Accurate Three-Dimensional Tracking Method in Bistatic Forward Scatter Radar,” URASIP
Journal on Advances in Signal Processing, no. 66, 2013, doi: 10.1186/1687-6180-2013-66.
[6] M. Cherniakov, M. Salous, V. Kostylev, and R. Abdullah, “Analysis of forward scattering radar for ground target detection,”
European Radar Conference, 2005. EURAD 2005., 2005, pp. 145-148, doi: 10.1109/EURAD.2005.1605586.
[7] M. Gashinova, V. Sizov, N. A. Zakaria, and M. Cherniakov, “Signal Detection in Multi-Frequency Forward Scatter Radar,” The
7th European Radar Conference, 2010, pp. 276-279. [Online]. Available: https://ieeexplore.ieee.org/document/5614901
[8] L. X. -Liang, H. Cheng, and Z. Tao, “The Analysis of Multi-polarization Forward Scattering RCS and the Effect on Target
Classification and Identification,” Journal of Electronics & Information Technology, vol. 32, no. 9, pp. 2191-2196, 2010,
doi: 10.3724/SP.J.1146.2009.01233.
[9] N. H. A. Aziz and R. S. A. R. Abdullah, “Speed Recognition Based PCA on Ground Vehicle in Passive Forward Scattering
Radar,” Journal of Telecommunication, Electronic and Computer Engineering, vol. 8, no. 3, pp. 133-137, 2016. [Online].
Available: https://core.ac.uk/download/pdf/229267341.pdf
[10] N. H. A. Aziz and M. A. Md Thani, “Vehicle Classification Using Passive Forward Scattering Radar,” Advanced Science Letters,
vol. 23, no. 11, pp. 11432-11436, 2017, doi: 10.1166/asl.2017.10299.
[11] N. F. Abdullah, N. E. A. Rashid, Z. I. Khan, and I. Musirin, “Analysis Of Different Z-Score Data to The Neural Network for
Automatic FSR Vehicle Classification,” IET International Radar Conference 2015, 2015, pp. 1-5, doi: 10.1049/cp.2015.1089.
[12] K. A. Othman, M. I. Jusoh, N. E. A. Rashid, and C. W. F. C. W. Fadhil, “Wavelet Technique Implementation in Forward
Scattering Radar (FSR) Ground Target Signal Processing,” Journal of Telecommunication, Electronic and Computer Engineering
(JTEC), vol. 9, no. 1-5, pp. 59-62, 2017. [Online]. Available: https://web.archive.org/web/20180417084300id_/
http://journal.utem.edu.my/index.php/jtec/article/viewFile/1835/1124
[13] C. Cao, J. Zhang, J. Meng, X. Zhang, and X. Mao, “Clutter Suppression and Target Tracking by the Low-Rank Representation for
Airborne Maritime Surveillance Radar,” in IEEE Access, vol. 8, pp. 160774-160789, 2020, doi: 10.1109/ACCESS.2020.3021124.
[14] H. Zeng, C. Hu, and T. Li, “The signal to noise ratio analysis of ground forward scattering radar,” 2009 IET International Radar
Conference, 2009, doi: 10.1049/cp.2009.0334.
[15] M. B. Porfido, M. Martorella, M. Gashinova, and M. Cherniakov, “Sea Clutter Power Reduction in Pulse Forward Scatter Radar,”
2018 19th International Radar Symposium (IRS), 2018, pp. 1-9, doi: 10.23919/IRS.2018.8448221.
[16] M. Gashinova, M. Cherniakov, N. A. Zakaria, and V. Sizov, “Empirical Model of Vegetation Clutter in Forward Scatter Radar
Micro-Sensors,” 2010 IEEE Radar Conference, 2010, pp. 899-904, doi: 10.1109/RADAR.2010.5494494.
[17] J. B. Billingsley, Low-angle Radar Land Clutter: Measurements and Empirical Models, William Andrew Publishing, 2002. [Online].
Available: https://books.google.co.id/books/about/Low_angle_Radar_Land_Clutter.html?id=FEkn0-h7sz0C&redir_esc=y.
[18] N. N. Ismail, N. E. A. Rashid, Z. I. Khan, N. E. A. Rashid, and M. F. Abdul Rashid, “Measurement, processing and modeling of a
tropical foliage clutter using Forward Scatter Radar micro-sensor network with VHF and UHF bands,” 2015 International
Conference on Radar, Antenna, Microwave, Electronics and Telecommunications (ICRAMET), 2015, pp. 76-81,
doi: 10.1109/ICRAMET.2015.7380778.
[19] L. Karray and A. Martin, “Towards improving speech detection robustness for speech recognition in adverse conditions,” Speech
Communication, vol. 40, no. 3, pp. 261-276, 2003, doi: 10.1016/S0167-6393(02)00066-3.
[20] P. Jancovic and M. Kokuer, “Incorporating the voicing information into HMM-based automatic speech recognition in noisy
environments,” Speech Communication, vol. 51, no. 5, pp. 438- 451, 2009, doi: 10.1016/j.specom.2009.01.003.
[21] D. Kumlu and I. Erer, “Clutter removal techniques in ground penetrating radar by using non-local means approach,” Journal of
the Faculty of Engineering and Architecture of Gazi University, vol. 35, no. 3, pp. 1269-1284, 2020. [Online]. Available:
https://dergipark.org.tr/en/download/article-file/1039916
[22] P. Jancovic, X. Zou, and M. Kokuer, “Speech Enhancement Based on Sparse Code Shrinkage Employing Multiple Speech
Models,” Speech Communication, vol. 54, no. 1, pp 108-118, 2012, doi: 10.1016/j.specom.2011.07.005.

Clutter reduction technique based on clutter model for … (Nur Emileen Abd Rashid)
962  ISSN: 1693-6930

[23] N. Smitha and V. Singh, “Clutter Reduction Techniques of Ground Penetrating Radar for Detecting Subsurface Explosive
Objects,” 2016 International Conference on Information Communication and Embedded Systems (ICICES), 2016, pp. 1-8,
doi: 10.1109/ICICES.2016.7518852.
[24] T. Pető and R. Seller, “Adaptive Clutter Cancellation Techniques for Passive Radars”, Topics in Radar Signal Processing,
Graham Weinberg, IntechOpen, 2017, doi: 10.5772/intechopen.71289.
[25] H. Kikuchi, E. Yoshikawa, T. Ushio, and Y. Hobara, “Clutter Reduction for Phased-Array Weather Radar Using Diagonal Capon
Beamforming with Neural Networks,” IEEE Geoscience and Remote Sensing Letters, vol. 17, no. 12, pp. 2065-2069, 2020,
doi: 10.1109/LGRS.2019.2962558.

BIOGRAPHIES OF AUTHORS

Nur Emileen Abd Rashid is currently a Senior Lecturer at University Technology

MARA. She obtained the Bachelor of Electrical Engineering (Telecommunication Engineering)
from the Universiti Kebangsaan Malaysia in 2001 and subsequently her M.Sc. in Computer,
Communication and Human Centered Engineering from the University of Birmingham, UK in
2002. She pursued her studies and received her PhD in 2011 from the same university. She has
contributed much of her expertise in areas related Radar Technology, Telecommunication signal
processing and clutter modelling. Dr. Nur Emileen is an active member of IET, IEEE and
MyRAN. She can be contacted at email: [email protected].

Nor Ayu Zalina Zakaria is currently a Senior Lecturer in School of Electrical

Engineering, College of Engineering, University Teknologi MARA, Malaysia and has almost 18
years working experience in an engineering education. She obtained her PhD degree in Radar
Technology from University of Birmingham, United Kingdom in 2017. She received her BEng
(Hons) in Electronics and Telecommunication Engineering from University Malaysia Sarawak and
Master in Mobile and Satellite Communication Engineering from University of Surrey, United
Kingdom in 2000 and 2003 respectively. Her areas of interest mainly in radar technology
particularly, radar detection and related work in clutter analysis for forward scatter radar. She can
be contacted at email: [email protected].

Zuhani Ismail Khan is an Electrical and Electronic Engineer, graduated from

University of Bradford, UK in 1998. After joining UiTM in the year 2000, she pursued her MSc in
Computer and Information Engineering at International Islamic University Malaysia and
completed it in 2005. She then completed her PhD in Electrical Engineering at Universiti
Teknologi MARA, UiTM in 2015 where she was involved in designing RF filters. Her research
background concentrates on Microwave and RF devices for high frequency system. On October
2019 she was appointed as the head of research in microwave research institute (MRI) and
currently she holds the post of Deputy Director of the institute. She can be contacted at email:
[email protected].

Siti Amalina Enche Ab Rahim received the Diplôme d’Ingénieur in electronics and
communications engineering from Ecole National Superieur d’Electronique et de Radioelectricite
de Grenoble (ENSERG), Grenoble, France in 2008 and received her D.Eng in electrical and
electronics engineering from Kyushu University, Japan in 2017. In 2009, she joined Telekom
Research & Development Sdn. Bhd (TMRND), Cyberjaya, Malaysia as a researcher, before
joining Universiti Teknologi Mara, (UiTM), Shah Alam, Malaysia as a lecturer in 2017. Her
current research interests include the design of RF passive components and RF CMOS integrated
circuits. She can be contacted at email: [email protected].

Nur Luqman Saleh is currently a research associate at Department of Computer and

Communication System Engineering, Faculty of Engineering, Universiti Putra Malaysia. He
received his Bachelor Degree in Electronic Engineering, Majoring in Microwave and
Communications in 2012 from the Multimedia University (MMU), Cyberjaya Malaysia. In 2014,
the author received his Master Degree of Engineering Management from Universiti Putra Malaysia
(UPM), Serdang, Malaysia. He completed his Ph.D. program in Communications and Network
Engineering at Universiti Putra Malaysia (UPM), Serdang, Malaysia in 2020. His research interest
related to human echolocator, radar applications, signal processing and data analytic. Prior
pursuing his post-graduate study, he was served as application support engineer at
telecommunications company. He can be contacted at email: [email protected].

TELKOMNIKA Telecommun Comput El Control, Vol. 20, No. 5, October 2022: 955-962