The part consists of two chapters. The first chapter is dedicated to exploring the fundamental principles of passive radar, including its history, the characteristics of opportunity signals, bistatic geometry, power budget, and passive radar coverage. The aim of this chapter is to provide readers with a comprehensive understanding of the theoretical foundations of passive radar.

The second chapter delves into the modeling of the received signals in the surveillance and reference channels. It then introduces two main categories of passive radar target detection techniques. Within this chapter, conventional approaches to passive radar signal processing are examined, where interference removal and target detection are performed as separate steps. These conventional methods rely on intuition rather than intrinsic optimality. The author stresses the importance of adopting detection theory-based approaches to effectively formulate the passive radar target detection problem in the presence of interference signals, including multipath and interfering target signals. Since opportunistic signals are not designed for radar applications and are beyond the control of radar designers, it becomes crucial to concentrate on devising more sophisticated and optimal signal processing algorithms for passive radar receivers. Together, these two chapters comprehensively cover diverse aspects of passive radar systems and serve as a valuable resource for researchers, engineers, and students interested in passive radar technology.

This book explores the problem of detecting targets using multistatic passive radar (PR) under the framework of detection theory. The term "multistatic" indicates that multiple transmitter-receiver pairs are used for target detection, with the receivers being spatially separated. In this type of radar, there is no dedicated transmitter, but instead, transmitters of opportunity, such as radio or television (TV) transmitters, are used.

In this chapter, we first model the received signal in the surveillance and reference of a bistatic passive radar. Then, we classify passive radar target detection techniques into two main categories, based on whether a reference channel is employed or not. In this chapter, we concentrate on the conventional signal processing approaches in passive radars and provide extensive simulation results to show that these methods are not optimal for the passive radar detection problem. Therefore, it is necessary to devise a target detection algorithm that considers the nature of opportunity signals.

Passive bistatic radar (PBR) systems commonly employ a reference channel (RC) to capture transmitted signal and a surveillance channel (SC) to receive echoes from the interested targets. In the first category of passive radar target detection approaches, it is assumed that the RC provides a high direct-path signal-to-noise ratio (SNR). This is often referred to as the ideal reference (IR) channel or high-SNR RC. This assumption allows for efficient target detection in the SC, even in the presence of the receiver noise, the direct-path signal, multipath/clutter echoes, and interfering targets. All target detection algorithms based on this assumption fall into the first category, known as Ca.1-IR approaches.

In this part, we provide a comprehensive review of detection-theory-based target detection approaches in Ca.1-IR. Specifically, this part is structured in the following way: Chapter 3 explores the problem of detecting multiple targets in a single-channel FM-based PBR. Chapters 4 and 5 investigate this problem in the context of multiband FM-based PBR to enhance target detection quality and target range resolution, respectively. In Chapter 5, a novel approach for multiband target detection is proposed, which simultaneously improves target detection quality and target range resolution. The multi-target detection problem in passive radar is then modeled as an M-ary hypothesis testing problem and presented in Chapters 6 and 7 for target detection in FM-, DVB-T-, and analog TV-based passive radars. Finally, Chapter 8 focuses on multi-accelerating-target detection problem of passive radars.

A systematic framework is presented in this chapter, which covers signal modeling, detection method, and statistical analysis, for bistatic PR target detection. In this chapter, a two-stage GLR detector has been developed to detect targets in the presence of interference, which includes noise, direct signal, clutter/multipath, and interfering targets. To detect targets in a multi-target scenario, a multistage algorithm has been proposed, which detects and eliminates the strongest target in order of decreasing signal strength. The false alarm probability and detection probability are calculated through closed-form expressions, which support the CFAR behavior and the detection performance loss of the proposed detector in the presence of interfering targets. Simulation results indicate that the proposed method is effective in controlling false alarm probability and achieving superior detection performance in the presence of noise, clutter, and interfering target signals. The detection quality of the multistage algorithm is shown to depend on the content of broadcasted FM signals, as demonstrated by the concept of DL. Next chapters will focus on the multiband detection problem to enhance target detection quality and target range resolution.

From our discussions, it is clear that detecting targets through PR involves intricate computational signal processing to achieve comparable performance to traditional active radar. However, we are fortunate that recent advancements in both digital signal processor hardware and algorithms have been made to help deal with these added complexities.

This chapter addresses the problem of detecting multiband single-input multiple-output (SISO) passive radar targets in the presence of interference signals, such as receiver noise, direct signal, multipath/clutter echoes, and interfering targets. The chapter presents a comprehensive framework that includes signal modeling, detection methods, and statistical analysis. Multiband SISO passive radar target detection is formulated as a binary composite hypothesis-testing problem, which is solved using the UMPI classical method. The false alarm probability and detection probability in the presence of targets with SW0 and SW1 models are derived in closed forms for the multiband UMPI test. This approach not only demonstrates that the proposed UMPI test has a CFAR property against noise variance but also provides robust performance against the time-varying program content of FM radio signals. Additionally, the multiband UMPI test benefits from combined diversity gain due to a combination of multifrequency bands and frequency diversity gain due to independent returns from one target compared with the single-band UMPI (GLR) test presented in Chapter 3.

This chapter proposes a signal processing method to enhance target range resolution, which involves a systematic framework comprising signal modeling, UMPI-based detection, and statistical analysis. The UMPI-based detection method is found to offer superior range resolution capabilities compared to the single-band detectors discussed in Chapter 3, as supported by theoretical analysis and simulations. The amount of range resolution improvement depends on the bandwidth of the processed multiband signal, but the quality of this improvement is influenced by the quality of the channels used. In some cases, reducing the bandwidth of individual channels may be preferred to improve the quality of target range resolution improvement. Additionally, the impact of complex amplitude mismatch over different channels is examined, revealing that it can result in reduced detection performance and worsen target range resolution improvement.

In this chapter, the focus was on addressing the issue of target detection in passive bistatic radar systems that use broadband FM. The chapter illustrates how the newly proposed broadband detection algorithm combines the advantages of the algorithms introduced in Chapters 4 and 5. The analytical framework developed in the chapter confirms that the proposed detector enhances both target range resolution and detection quality. Simulation results provide evidence that the algorithm is effective.

This chapter presents a model for detecting targets in PR systems using an M-ary hypothesis test for multitarget scenarios. The proposed approach is a forward and sequential GLR-based detector that detects targets one at a time and treats previously detected targets as interferences, enabling detection of even weak targets. This chapter also proposes a parallel implementation of the GLR-based detector to reduce memory requirements. Closed-form expressions for threshold and detection probability are derived. Simulation results show that the proposed algorithm outperforms existing methods without significant added complexity.

The signal used in ATV-based PBR is not under the control of radar system designers, so the only solution is to manage the signal's range ambiguities on the receiver side. This requires a sophisticated algorithm to prevent false alarms and target masking, particularly in the range dimension, and to estimate ranges in a multitarget scenario. To do this, a multilayer detector has been proposed in which one target is detected in each layer based on the GLR detector. Interference due to the detected target in the first layer was estimated and subtracted from the received signal. From the first layer's interference-canceled signal, another target (if it exists) was detected in the second layer. These detection and interference cancellation steps were carried out in each layer until all the targets were detected. To reduce the computational complexity, we use a chirp z-transform-based algorithm to compute the GLR-based detector over a limited frequency range. We also present a robust target detection algorithm to effectively cancel strong interference and off-grid targets. Our simulations demonstrate that the proposed detection algorithms can jointly estimate the number and delay-Doppler coordinates of desired targets.

First, we have investigated the problem of detecting a maneuvering target using the UMPI detector designed in Chapter 4. The closed-form expression for the detection probability of the UMPI detector has been obtained in the presence of a maneuvering target. Simulation results show that the bistatic velocity and acceleration of a maneuvering target induce range migration and Doppler frequency migration of echo signal, which degrade the performance of the UMPI detector in the integration time as long as seconds. To solve this, then, we proposed a new 3-dimensional detector in which delay-Doppler and acceleration of interesting targets are estimated, so it was called a 3-dimensional detection algorithm. To reduce the computational complexity of the proposed GLR-based detector, it was implemented based on a modified version of the chirp fast Fourier transform. To reduce its computational complexity, a 3-dimensional sequential algorithm, in which targets have been detected in a layered manner, was proposed for multitarget scenarios. A closed-form formula for the false alarm probability of the proposed detector has been derived which is very instructive to adjust the detection threshold. Our simulation results indicated that the 3-dimensional detection algorithm is required to prevent degraded detection probability as well as excessive false alarm probability of the 2-dimensional classical target detection algorithm, especially for highly accelerating targets.

Passive radars utilize signals from non-cooperative sources like radio, television, and cellular signals instead of having their own dedicated transmitter to detect targets of interest. Traditional matched filter-based detectors used in active radars may not be suitable for passive radars because the transmitted signals are typically unknown to the passive radar receiver. One solution is to exploit an additional channel, known as a reference channel, to collect a delayed version of the unknown transmitted signal. However, many existing methods assume a high direct-path signal power-to-noise ratio (DNR) in the reference channel and ignore the noise. These methods are categorized as category 1 with an ideal reference channel (Ca.1-IR) and have been extensively studied in Part 2. However, their performance can be significantly impacted, especially when the DNR of the reference channel is low. In this part, we explore situations in which it is not possible to ignore the effects of noise in the reference channel. Consequently, we encounter a challenge known as passive radar target detection with a noisy reference (NR) channel. Similar to the Ca.1-IR methods, here, we employ two receive channels for target detection: one for surveillance and the other for reference. This similarity leads us to categorize the target detection techniques developed for the NR channel as Ca.1-NR methods. Although both Ca.1-IR and Ca.1-NR detection strategies fall under the same category of passive radar target detection methods, they differ in hypothesis-testing problem formulations. Firstly, Ca.1-NR approaches incorporate both reference and surveillance channels when formulating the passive radar target detection problem, while Ca.1-IR methods only use the surveillance channel for this purpose. Secondly, in Ca.1-NR approaches, it is not feasible to treat multipath signals as interference; thus, these interference signals are ignored to obtain closed-form detectors. However, it is possible to utilize the reference channel separately to estimate the unknown transmitted signal and exploit the subspace-based interference removal algorithm of Chapter 7 to cancel the effects of multipath signals.

This chapter introduces a new method for setting thresholds in radar applications. Rather than fixing the size of a test, we propose fixing the level of the test for practical uncertainty conditions. We present three new detectors based on detection theory and the combination of detection theory and kernel theory. The DNR _r of the reference channel has a significant impact on the detection thresholds and false alarm regulations of the proposed detectors when using the noisy reference channel. To address this, we suggest setting the detection thresholds of NR-based detectors so that their levels are fixed according to a desired false alarm probability. In our proposed kernel theory-based detectors, we improve detection performance by replacing the inner products of test statistics with appropriate polynomial kernel functions. We use the principle of invariance and maximizing detection performance to set the polynomial kernel parameters of the proposed kernelized detectors. Through extensive Monte Carlo simulations, we demonstrate the effectiveness of our threshold-setting strategy and show that the proposed kernelized detectors offer over 1 dB of detection performance gain compared to their counterparts.

In this chapter, two problems related to detecting targets in a multistatic setting have been formulated: one involving two channels and another involving a single channel. Two detectors, named P1-LRT and P2-LRT, have been developed to solve these problems using the LRT criterion. It has been observed that the DNR_avg of the RC has a significant impact on the detection thresholds and false alarm regulation of the P1-LRT detector. To address this issue, a strategy has been proposed to set the detection threshold of the P1-LRT detector to achieve a fixed level of desired false alarm probability. Despite being a fixed-size test, the performance of the P2-LRT detector is not as good as that of the P1-LRT detector. MC simulation examples have been used to demonstrate the effectiveness of the proposed threshold-setting strategy and to show that the P1-LRT detector outperforms the P2-LRT detector in low and high DNR_avg regimes.

This part deals with detecting passive radar targets in situations where a reliable reference channel is not available due to various factors, such as low signal-to-noise caused by obstructions in the direct path between transmitters and receivers or strong multipath clutter. In such cases, incorporating both the reference and surveillance channels to formulate the target detection problem may lead to inadequate performance of the designed detectors in detecting the targets. Alternatively, one can improve performance by removing the reference channel and utilizing multiple surveillance channels, which allows for more data to be collected and provides spatial diversity for a target’s radar cross section. Additionally, interchannel correlations among different receivers can be utilized for target detection and estimating unknown transmitter signals. This approach falls under the second category of passive radar target detection methods, specifically Category 2 (Ca.2). This category itself can be considered a special case within the first category (Ca.1), where only the surveillance channel is included to formulate the target detection problem.

This chapter has unified the target detection problem in an uncalibrated multistatic passive radar with distributed receivers. To achieve this, we formulated and solved three multistatic target detection problems using the Rao, geometric argument, and LRT principles, resulting in four new detectors. Moreover, we carried out extensive MC simulations to evaluate the performance of the proposed detectors and found that each one has its own unique capabilities. In addition, we utilized the invariance theory to examine the potential CFAR behavior of the detectors against noise variance uncertainties across different receivers. Our analysis showed that, with the exception of the P₁-Vol detector, all the proposed detectors exhibit robustness against noise variance uncertainties in uncalibrated receivers.

This chapter has provided details to develop the Rao test for detecting targets in calibrated/uncalibrated passive radar systems. These detectors are advantageous compared to traditional LRT approaches because they are in closed form, eliminating the need for iterative algorithms of LRT-based detectors. This makes them more practical. The chapter provides several MC simulation results to demonstrate the effectiveness of the proposed Rao-based detectors in comparison to the detector discussed in Chapter 12. The results indicate that the proposed detectors outperform their counterparts for various multistatic configurations, particularly in terms of detection probability and false alarm regulation.