# Analysis on Target Detection and Classification in LTE Based Passive Forward Scattering Radar

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Target Detection in Passive Forward Scattering Radar

#### 2.1. Passive Forward Scatter Radar Cross Section

#### 2.2. Target Doppler Signature in Passive FSR

_{dr}, (assume phase noise free) and a scattered signal from a moving target, S

_{sc}carrying a Doppler shift, ω

_{sc}[2,13] as depicted in Equations (1) and (2) respectively. In a simplified manner, the received signal at the input of the passive FSR system S

_{rx}can be expressed as Equation (3):

_{dr}represents the amplitudes of the direct transmission signal (leakage) and A

_{sc}is the amplitude of the signal scattered from the moving target. As stated earlier, the system uses a non-linear element as its detection mechanism with the transfer characteristics of Y

_{out}= (Y

_{in})

^{2}, hence at the output of the non-linear element and in this case the self-mixing between S

_{dr}and S

_{sc}, the signal could be represented by:

_{out}; the direct path signal (leakage signal) and the signal with Doppler frequency scattered from the target. The filtered signal is denoted as Y

_{out_f}and shown in Equation (5):

_{out_f}= A

_{dr}A

_{sc}sin(ω

_{sc}t). The total Doppler frequency can be extracted by deriving the phase component of Equation (5): and is given by:

#### 2.3. Received Signal and Target’s Features in Passive FSR

- the amplitudes of Doppler signatures to the target’s forward scatter cross section, hence, the Saloon car had a bigger forward scatter cross section, which was reflected in the amplitudes of their signature,
- the signal span in time domain to the length of the target, hence, the Saloon car also had a longer body, which was reflected in the signal span of the forward scatter main lobe.

## 3. Target Classification in Passive Forward Scattering Radar

**V**

^{c}}

_{i}formed for each vehicle category. Feature vectors {

**V**

^{c}}

_{i}is saved and will be used as a reference during the classification process. In the classification stage, the passive FSR system will capture an unknown signal from vehicle crossing the baseline. The new signal is converted into the feature vectors set for individual vehicle. By using the same transformation matrix H, it is then transformed to the feature vector

**V**

_{j}and is placed into the PCA space together with Feature vectors {

**V**

^{c}}

_{i}. The final step in classification stage is to apply classifier rule and algorithm to the new signal vehicles’ feature vectors,

**V**

_{j}which is unknown. There are many choices of the established classifier method can be applied. The simplest yet reliable classifier method, which is K Nearest Neighbors (KNN) with Euclidean distance, is used. Here three neighbors were chosen to find the smallest distance before category decision was made by the system.

#### 3.1. Evaluation of Target’s Baseline Crossing

_{R}(distance between vehicle to the receiver) in passive FSR can be predicted by using the same system as shown in Figure 8. The effect of vehicle’s speed to the overall precision is also analyzed. The variance explained from the output of PCA is utilized for the clustering purposes. In statistics, the explained variation measures the proportion of the mathematical model for the variation of a given dataset. Frequently, variation is quantified as variance, where only then the more specific term ‘variance explained’ can be used [22]. Figure 10 shows the PCA variance explained of the training data for target speeds of 5, 10, 20 and 30 km/h in the PCA space. The results are summarized in Table 4, which highlights the experimental results of the variance explained on the training data. In the experiment, 120 samples were used for each speed of the moving target. Three types of target distance from the passive radar receiver (baseline) were to be recognized by the clustering-based PCA (5, 10 and 20 m). For every type of target’s baseline with the same speed, 40 samples were used. Overall, 480 samples were used for the training data in the principal component space expressed by the first two principal components. Finally, the higher speed of the moving target contributes a finer variance explained, such as 82% of the variance of the training data for 30 km/h, which is the highest among the moving target’s variation of speed, where it is proficiently used for the first principal component. However, for the speeds of 10 km/h and 20 km/h, the variance explained were 63% and 48% respectively as the PCA was built from components such as the sample covariance, which are not statistically robust. This meant that the PCA may be thrown off by outliers and other data pathologies. How seriously this affects the results is specific to the data and application [22].

#### 3.2. Vehicle Classification Performance at Different Baseline Crossing Range

_{R}(range from vehicle to the receiver) affect the vehicle classification performance does. It is known that from the FSR theory and as also discussed in Section 2, when a moving target crossing the baseline, the system will gain the enhancement in FS RCS, ${\sigma}_{f}$ and is given by:

_{T}is the LTE transmitted power from the base station, G

_{T}is the LTE transmitter antenna gain, G

_{R}is the receiver antenna gain. The range from the transmitter to the target is given by R

_{T}. The received signal is highly dependent onthe target size and range between the target and receiver. This data could reflect the signal power at the antenna. The received signal power is inversely proportional to the R

_{R}by the inverse square law. However, the range between the target and receiver is restricted in ensuring the FSR condition is met. Figure 12 shows the time domain signal scattered from vehicle in the Saloon category for different R

_{R}with v = 10 km/h. Obviously, this signal is the output after all the hardware processing including first RF amplification, amplitude detection via diode, high order frequency filtering by LPF, DC blockage by HPF and low frequency amplifier. It can be seen that the amplitude envelope decreases as the range, R

_{R}increasing. As the system is linear, the envelope does also represent the actual signal power pattern at the front end of the receiver. This signal is then used as the input to the classification system explained in Figure 8. The overall classification performance is tabulated in Table 5 and the location of training and testing data in the PC space is illustrated in Figure 13. As expected, highest classification is observed for R

_{R}= 5 m, which follows the previous indicator explained in Section 3.1. The performance is slightly decreases as the range R

_{R}increases which reflect to the quality of the input signal to the classification system. It can be suggested that signal strength plays important role in the classification performance.

## 4. Conclusions

## Author Contributions

## Conflicts of Interest

## Abbreviations

FSR | Forward Scattering Radar |

RCS | Radar Cross Section |

LTE | Long-Term Evolution |

AF | Ambiguity Function |

FSML | Forward Scatter Main Lobe |

FS | Forward Scatter |

GSM | Global Systems for Mobile |

SUV | Sport Utility Vehicle |

CST | Computer Simulation Technology |

ADC | Analogue to Digital Converter |

PCA | Principle Components Analysis |

FFT | Fast Fourier Transform |

PC | Principle Component |

## References

- Cherniakov, M. Bistatic Radar: Emerging Technology; John Wiley and Sons, Ltd.: Chichester, UK, 2008. [Google Scholar]
- Cherniakov, M.; Abdullah, R.S.A.R.; Jancovic, P.; Salous, M.; Chapursky, V. Automatic ground target classification using forward scattering radar. IEE Proc. Radar Sonar Navig.
**2006**, 153, 427–437. [Google Scholar] [CrossRef] - Marra, M.; De Luca, A.; Hristov, S.; Daniel, L.; Gashinova, M.; Cherniakov, M. New algorithm for signal detection in passive FSR. In Proceedings of the 2015 IEEE Radar Conference, Johannesburg, South Africa, 27–30 October 2015; pp. 218–223.
- Pastina, D.; Contu, M.; Lombardo, P.; Gashinova, M.; De Luca, A.; Daniel, L.; Cherniakov, M. Target motion estimation via multi-node forward scatter radar system. IET Radar Sonar Navig.
**2016**, 10, 1. [Google Scholar] [CrossRef] - Griffiths, H.D.; Long, N.R.W. Television-based bistatic radar. IEE Proc. F Commun. Radar Signal Process.
**1986**, 133, 649–657. [Google Scholar] [CrossRef] - Howland, P.E.; Maksimiuk, D.; Reitsma, G. FM radio based bistatic radar. IEE Proc. Radar Sonar Navig.
**2005**, 152, 107–115. [Google Scholar] [CrossRef] - Poullin, D. Passive detection using digital broadcasters (DAB, DVB) with COFDM modulation. IEE Proc. Radar Sonar Navig.
**2005**, 152, 143–152. [Google Scholar] [CrossRef] - Bárcena-Humanes, J.L.; Gómez-Hoyo, P.J.; Jarabo-Amores, M.P.; Mata-Moya, D.; Del-Rey-Maestre, N. Feasibility study of EO SARs as opportunity illuminators in passive radars: PAZ-based case study. Sensors
**2015**, 15, 29079–29106. [Google Scholar] [CrossRef] [PubMed] - Griffiths, H.D.; Garnett, A.J.; Baker, C.J.; Keaveney, S. Bistatic radar using satellite-borne illuminators of opportunity. In Proceedings of the Radar International Conference, Brighton, UK, 12–13 October 1992; pp. 276–279.
- Hongbo, S.; Tan, D.K.P.; Lu, Y. Aircraft target measurements using A GSM-based passive radar. In Proceedings of the 2008 IEEE Radar Conference, Rome, Italy, 26–30 May 2008; pp. 1–6.
- Raja Abdullah, R.S.A.; Salah, A.A.; Ismail, A.; Hashim, F.H.; Abdul Rashid, N.E.; Abdul Aziz, N.H. Experimental investigation on target detection and tracking in passive radar using long-term evolution signal. IET Radar Sonar Navig.
**2016**, 10, 577–585. [Google Scholar] [CrossRef] - Raja Abdullah, R.S.A.; Rasid, M.F.A.; Mohamed, M.K.H. Improvement in detection with forward scattering radar. Sci. China Inf. Sci.
**2011**, 54, 2660–2672. [Google Scholar] [CrossRef] - Gashinova, M.; Daniel, L.; Hoare, E.; Sizov, V.; Kabakchiev, K.; Cherniakov, M. Signal characterisation and processing in the forward scatter mode of bistatic passive coherent location systems. EURASIP J. Adv. Signal Process.
**2013**, 2013, 36. [Google Scholar] [CrossRef] - Suberviola, I.; Mayordome, I.; Mendizabal, J. Experimental results of air target detection with GPS forward scattering radar. IEEE Geosci. Remote Sens. Lett.
**2012**, 9, 47–51. [Google Scholar] [CrossRef] - Behar, V.; Kabakchiev, C. Detectability of Air Target Detection using Bistatic Radar Based on GPS L5 Signals. In Proceedings of the 12th International Radar Symposium (IRS), Leipzig, Germany, 7–9 September 2011; pp. 212–217.
- Kabakchiev, C.; Garvanov, I.; Behar, V.; Daskalov, P.; Rohling, H. Study of moving target shadows using passive Forward Scatter radar systems. In Proceedings of the 2014 15th International Radar Symposium (IRS), Gdansk, Poland, 16–18 June 2014; pp. 1–4.
- Krysik, P.; Kulpa, K.; Samczyński, P. GSM based passive receiver using forward scatter radar geometry. In Proceedings of the 2014 15th International Radar Symposium (IRS), Dresden, Germany, 19–21 June 2013; pp. 637–642.
- Salah, A.A.; Abdullah, R.S.A.; Ismail, A.; Hashim, F.; Leow, C.Y.; Roslee, M.B.; Rashid, N.A. Feasibility study of LTE signal as a new illuminators of opportunity for passive radar applications. In Proceedings of the 2013 IEEE International RF and Microwave Conference (RFM), Penang, Malaysia, 9–11 December 2013; pp. 258–262.
- Ahmadi, S. An overview of 3GPP long-term evolution radio access network. In New Directions in Wireless Communications Research; Springer: Hillsboro, TX, USA, 2009; pp. 431–465. [Google Scholar]
- Duda, R.O.; Hart, P.E.; Stork, D.G. Pattern Classification; John Wiley & Sons: New York, NY, USA, 2000; pp. 2–4. [Google Scholar]
- Welch, P.D. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. IEEE Trans. Audio Electroacoust.
**1967**, 15, 70–73. [Google Scholar] [CrossRef] - Noor Hafizah, A.A.; Haziq Hazwan, M.Y.; Nur Emileen, A.R.; Raja Syamsul Azmir, R.A.; Asem, S. RCS analysis on different targets and bistatic angles using LTE frequency. Int. J. Ind. Electron. Electr. Eng.
**2015**, 3, 1–4. [Google Scholar]

**Figure 7.**Received scattered signal in (

**i**) time domain; (

**ii**) power spectrum and (

**iii**) spectrogram of moving vehicles crossing the Tx-Rx radar baseline for Compact and Saloon.

**Figure 9.**(

**a**) Location of training data in PCA space; and (

**b**) the spreading of testing data in the database of PC domain.

**Figure 10.**Variance explained of training data for ground moving target with speed of (

**a**) 5 km/h; (

**b**) 10 km/h; (

**c**) 20 km/h and (

**d**) 30 km/h in the PC space.

**Figure 11.**Location of training data for ground moving target with speed of (

**a**) 5 km/h; (

**b**) 10 km/h; (

**c**) 20 km/h and (

**d**) 30 km/h in the PCA space.

**Figure 12.**Signal in time at the ADC output for vehicle crossing at different crossing range, R

_{R}.

**Figure 13.**Location of training and testing data for all vehicle categories at different R

_{R}in PCA space for (

**a**) R = 5 m; (

**b**) R = 10 m and (

**c**) R = 20 m.

Parameters | Incident Angle, φ | |
---|---|---|

59° | 90° | |

Bistatic angle, β of Max RCS (°) | 180.00 | 180.00 |

Max RCS magnitude (dBm^{2}) | 60.04 | 62.75 |

Side lobe level (dB) | −31.30 | −23.50 |

Category | Compact | Saloon | Small SUV |
---|---|---|---|

Actual car used during experiment to represent each category | |||

No. of collected data used for classification | 70 | 73 | 70 |

Vehicle’s Dimension | L = 3395 mm | L = 4360 mm | L = 4420 mm |

H = 1415 mm | H = 1385 mm | H = 1740 mm |

Vehicle-Category | NO.V—For Testing | Automatically Classified as (%) | ||
---|---|---|---|---|

Compact | Saloon | Small SUV | ||

Compact | 30 | 92 | 0 | 8 |

Saloon | 33 | 0 | 100 | 0 |

Small SUV | 30 | 4 | 0 | 96 |

Target’s Speed (km/h) | 5 | 10 | 20 | 30 |

Variance Explained (%) | 73 | 63 | 48 | 82 |

Range, R_{R} (m) | 5 | 10 | 20 | ||||||
---|---|---|---|---|---|---|---|---|---|

Type of Vehicle | Compact | Saloon | SUV | Compact | Saloon | SUV | Compact | Saloon | SUV |

Training Data | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 |

Testing Data | 30 | 69 | 33 | 27 | 61 | 35 | 32 | 39 | 25 |

Classification | 30 | 67 | 33 | 21 | 57 | 32 | 30 | 33 | 17 |

Classification (%) | 100 | 97.1 | 100 | 77.78 | 93.44 | 91.4 | 93.75 | 84.62 | 68.0 |

Average Classification (%) | 99.0 | 87.54 | 82.12 |

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Raja Abdullah, R.S.A.; Abdul Aziz, N.H.; Abdul Rashid, N.E.; Ahmad Salah, A.; Hashim, F.
Analysis on Target Detection and Classification in LTE Based Passive Forward Scattering Radar. *Sensors* **2016**, *16*, 1607.
https://doi.org/10.3390/s16101607

**AMA Style**

Raja Abdullah RSA, Abdul Aziz NH, Abdul Rashid NE, Ahmad Salah A, Hashim F.
Analysis on Target Detection and Classification in LTE Based Passive Forward Scattering Radar. *Sensors*. 2016; 16(10):1607.
https://doi.org/10.3390/s16101607

**Chicago/Turabian Style**

Raja Abdullah, Raja Syamsul Azmir, Noor Hafizah Abdul Aziz, Nur Emileen Abdul Rashid, Asem Ahmad Salah, and Fazirulhisyam Hashim.
2016. "Analysis on Target Detection and Classification in LTE Based Passive Forward Scattering Radar" *Sensors* 16, no. 10: 1607.
https://doi.org/10.3390/s16101607