Exact Detection Probability for Partially Correlated Rayleigh Targets

The detection probability PD of a radar receiver which postdetection integrates N pulses of an expqnentially correlated signal from a Rayleigh target in thermal noise is determined. At the limiting correlation coefficients, p = 1 and p = 0, the analysis yields, respectively, the well known Swerling case 1 and case 2 formulas. The effect of partial (0 ? p ? 1) correlation is exhibited in a set of curves of PD versus signal-to-noise ratio, X, for various N and p. Additional curves compare the exact fluctuation loss determined from the above analysis with an approximate expression universally employed by radar system engineers.     

Two-Stage CFAR Detectors for Multiple-Range-Bin Radars

Two-stage detectors using generalized sign and four-level conditional statistics for signal detection in multiple-range-bin radars are described. The resulting detectors are of constant false-alarm rate (CFAR). Performances are evaluated and compared with singlestage versions. If the a priori probability of either the no-target case or the target-presence case is large (close to 1), a two-stage test can be designed to have the advantage of reducing the average number of samples required without sacrificing detection probability. With the proper choices of parameters, significant improvement in the efficiency can be achieved. Asymptotic relative efficiency of two-stage detectors with respect to single-stage detectors is derived and some numerical results are evaluated.     

Envelope Threshold Detection in a Class of Non-Gaussian Interference

This paper considers the detection of a sinusoidal or chirp signal imbedded in wideband FM interference (as might be generated by some types of active jamming), such that after pulse compression or other integration, the interference can be approximated by a sum of sinusoids of independent phase. The detection probability in such non-Gaussian noise is compared to that for Gaussian noise, with the Gaussian result approached, as required, in the limit that the number of sinusoids in the sum increases without bound. For detection using a comparison of the envelope with a threshold which yields a given false-alarm probability (CFAR detection), the detection probability is improved over the case of Gaussian noise, so that the usual approach basing the design on Gaussian noise would be conservative. Using a threshold determined from the envelope mean, the FM interference yields a lower false-alarm probability than for Gaussian noise, with detection probability only slightly degraded.     

A noncoherent adaptive detection technique

An adaptive detection technique suitable for both stationary and nonstationary noise environments based upon a generalized likelihood ratio test (GLRT) formulation is presented. The detector, which is statistically equivalent to a special form of the Wilks's lambda test, noncoherently combines the information contained in a pulse train of arbitrary length for decision-making purposes. The probability density function of the test under the noise only hypothesis is shown to be central χ². Under the signal plus noise hypothesis, an exact statistical characterization of the test cannot be obtained, and, therefore, a Chernoff bound is derived. Results in terms of the probability of detection versus signal-to-noise ratio (SNR) obtained from Monte Carlo simulation, the Chernoff bound, and the optimal matched filter case are examined. The performance of the noncoherent detector is shown to be a function of the covariance matrix estimate and the number of data samples     

Interference Effects of Pseudo-Random Frequency-Hopping Signals

When a pseudo-random frequency-hopping signal is intercepted by a conventional receiver operating within the same frequency band, the interfering signal has the form of a pulse-amplitude modulated signal. Each pulse amplitude is dependent upon the hopping frequency and the selectivity characteristic of the victim receiver. The probability density function for the interfering pulse amplitude prior to demodulation is determined when the probability density function for the hopping frequency is uniform and the victim-receiver characteristic is 1) ideal flat bandpass, 2) single tuned, and 3) Gaussian shaped. It is shown that the average interfering pulse amplitude and interference power decrease as the frequency-hopping bandwidth increases with respect to the victim-receiver bandwidth. Fast Fourier transform computer techniques are used to obtain the probability density function of the interference amplitude in a Gaussian receiver when several (from 2 to 10) pseudo-random frequency-hopping systems are simultaneously using the same frequency band. The probability that the interference exceeds a prescribed threshold value is computed from the derived probability density functions. This probability may be used in signal-to-interference ratio calculations, to describe the capture effect, or to compute the expected number of clicks produced in an FM discriminator.     

Effects of two-way decorrelation on radar detection inscintillation

A 3 dB gain in average signal-to-noise ratio of a monostatic radar operating in scintillation has recently been established both theoretically and observationally. The statistics of two-way scintillation are derived here for the case where the uplink and downlink both experience Rayleigh fading and where there is arbitrary correlation between the scintillation on the two paths. These statistics are then used to compute radar detection curves. A surprising result is obtained. The probability of detection is only weakly dependent (for P _D in the range 0.1 to 0.9) on the degree of uplink-downlink correlation in the scintillation when the average (nonfading) signal-to-noise ratio is constant and when proper account is taken of the change in mean power between the monostatic and bistatic cases. Much larger differences are seen in the detection curves with scintillation compared with nonfading curves (for P_D equal to 0.7 this scintillation loss is about 7 dB). Thus the difference in detection performance of monostatic and bistatic radars is determined primarily by the difference in the radar cross section (RCS) of the target for the two cases     

Detection Probabilities for Log-Normally Distributed Signals

The amplitude and power of a large family of radio signals are observed to have log-normal probability density functions. Among these are signals propagated through random inhomogeneous media, a notable example being low frequency atmospheric radio noise. Of greater importance are certain radar targets that have been observed to have essentially log-normal density functions. Both ships and space vehicles may fall into this category. Curves of probability of detection vs. signal-to-noise ratio for the case of log-normal signals in Gaussian noise have been computed and are presented in this paper. The curves apply for square-law detection with varying degrees of postdetection linear integration. Both fully correlated and completely uncorrelated fluctuating signals are considered. It is shown that for log-normal signal distributions having large variances, the probability of detection differs significantly from that obtained using curves based on an assumed Rayleigh signal distribution.     

Detectability of Spread-Spectrum Signals

Methods of detecting spread-sprectrum signals without knowledge of the pseudorandom code used to generate the signal are described. Exact and approximate methods of calculating relationships among detection probability, false alarm rate, and signal-to-noise ratio are given for radiometers and for channelized pulse-detection systems. The detection performance of the radiometer is compared graphically with that of pulse-detection systems, for two different kinds of pulse detection decision rules. Detection performance as a function of certain signal parameters is shown to be very different for a pulse-detection system than for a radiometer, and this difference in behavior provides a basis for selecting signal parameters that minimize the probability of detection. The reasoning that underlies the selection process is explained, and the process is outlined for each of several signal parameters.     

Filter Selection for Receivers Using Square-Law Detection

A numerical method is described for predicting the detection probability performance of a pulse receiver which uses square-law detection. The method is useful for receivers where the ratio of RF bandwidth to video bandwidth is in the range from 2 to 40; a range where numerical results have previously been hard to obtain. A key feature of the approach is that it takes into account the actual filter transfer functions, pulse envelope shape, and pulse frequency.     

Calculation of Detection Probability for Frequency Compressive Receivers

A numerical method is described for predicting the detection probability of a frequency compressive pulse receiver. The approach is general; i.e., input pulse envelope shapes, pulse offset frequencies, and filter magnitude responses are arbitrary. The performance of the frequency compressive receiver is analyzed and compared to that of a crystal video receiver for a specific case.     

Error Probability in Binary Digital Communication in the Presence of an Interfering Signal

The presence of an interferring signal i(t) in a binary coherent symmetric channel is investigated. The probability of error is maximum when i(t) is antipodal to the transmitted signal, so(t) or s1 (t). The error probability is derived in the presence of a delayed interfering signal - so (t - r), where r is the time delay. An example of phase shift keying signals is given.     

Sidelobe blanking with integration and target fluctuation

In radar systems, sidelobe blanking (SLB) is used to mitigate impulsive interference that enters the radar through sidelobes of the main antenna. SLB employs an auxiliary antenna channel with the output being compared with that of the main antenna channel and a decision is then made as to whether or not to blank the main channel output. SLB performance determination involves the evaluation of several probability functions. Based on the classical Maisel SLB architecture, this work extends previous performance results, in which detection was limited to the case of a single radar pulse with either Marcum or Swerling I target fluctuation. Probability expressions have been generalized to include both an arbitrary number of integrated pulses and target fluctuation models based on the gamma distribution. The Swerling fluctuation models are all special cases of the gamma distribution. Results are derived in terms of two generalized probability functions, one for detection and the other for blanking. With these generalized probability functions, the SLB design and performance results can be determined. Examples are presented and discussed.     

Contiguous pulse binary integration analysis

The probability of detecting m or more pulses contiguously-that is, in a row-from a pulse train of n pulses is determined when the detection of each pulse is an independent Bernoulli trial with probability p. While a general closed-form expression for this probability is not known, we present an analytical procedure that gives the exact expression for the probability of interest for any particular case. We also present simple asymptotic expressions for these probabilities and develop bounds on the probability that the number of pulses that must be observed before m contiguous detections is greater than or less than some particular number. We consider the implications for binary integration in radar and electronic warfare problems     

Statistical Analysis of Wide-Band Pseudorandom Matched Filter Sonars

A stochastic model of pseudorandom (PR) signals is adopted and a statistical analysis carried out of the velocity (i. e., Doppler) and acceleration tolerances of wide-band PR matched filter sonar systems. The reference functions for these correlation detection systems are considered to be time-compressed or time-expanded replicas of the transmitted signal. Results are derived for the case of a PR signal having a flat power spectral density over a finite bandwidth. It is shown that the velocity and acceleration tolerances are essentially independent of the signal bandwidth and that therefore the radar-derived expressions for the narrow-band tolerances can be extended to the wide-band sonar case. An interesting result is that the derived acceleration tolerance is approximately three times the widely used estimate that is based on the target remaining in the same Doppler channel over the integration time.     

Energy Detection of a Random Process in Colored Gaussian Noise

A likelihood receiver for a Gaussian random signal process in colored Gaussian noise is realized with a quadratic form of a finite-duration sample of the input process. Such a receiver may be called a "filtered energy detector." The output statistic is compared with a threshold and if the threshold is exceeded, a signal is said to be present. False alarm and detection probabilities may be estimated if tabulated distributions can be fitted to the actual distributions of the test statistic which are unknown. Gamma distributions were fitted to the conditional probability densities of the output statistic by equating means and variances, formulas for which are derived assuming a large observation interval. A numerical example is given for the case in which the noise and signal processes have spectral densities of the same shape or are flat. The optimum filter turns out to be a band-limited noise whitener. The factors governing false alarm and detection probabilities are the filter bandwidth, the sample duration, and the signal level compared to the noise. Two sets of receiver operating characteristic curves are presented to complete the example.     

CFAR detection in clutter with unknown correlation properties

We develop a constant false-alarm rate (CFAR) approach for detecting a random N-dimensional complex vector in the presence of clutter or interference modeled as a zero mean complex Gaussian vector whose correlation properties are not known to the receiver. It is assumed that estimates of the correlation properties of the clutter/interference may be obtained independently by processing the received vectors from a set of reference cells. We characterize the detection performance of this algorithm when the signal to be detected is modeled as a zero-mean complex Gaussian random vector with unknown correlation matrix. Results show that for a prescribed false alarm probability and a given signal-to-clutter ratio (to be defined in the text), the detectability of Gaussian random signals depends on the eigenvalues of the matrix R_c^-1R_s. The nonsingular matrix R_c and the matrix R_s are the correlation matrices of clutter-plus-noise and signal vectors respectively. It is shown that the “effective” fluctuation statistics of the signal to be detected is determined completely by the eigenvalues of the matrix R_c^-1R_s. For example the signal to be detected has an effective Swerling II fluctuation statistics when all eigenvalues of the above matrix are equal. Swerling I fluctuation statistics results effectively when all eigenvalues except one are equal to zero. Eigenvalue distributions between these two limiting cases correspond to fluctuation statistics that lie between Swerling I and II models     

Maximum likelihood track-before-detect with fluctuating targetamplitude

An important problem in target tracking is the detection and tracking of targets in very low signal-to-noise ratio (SNR) environments. In the past, several approaches have been used, including maximum likelihood. The major novelty of this work is the incorporation of a model for fluctuating target amplitude into the maximum likelihood approach for tracking of constant velocity targets. Coupled with a realistic sensor model, this allows the exploitation of signal correlation between resolution cells in the same frame, and also from one frame to the next. The fluctuating amplitude model is a first order model to reflect the inter-frame correlation. The amplitude estimates are obtained using a Kalman filter, from which the likelihood function is derived. A numerical maximization technique avoids problems previously encountered in “velocity filtering” approaches due to mismatch between assumed and actual target velocity, at the cost of additional computation. The Cramer-Rao lower bound (CRLB) is derived for a constant, known amplitude case. Estimation errors are close to this CRLB even when the amplitude is unknown. Results show track detection performance for unknown signal amplitude is nearly the same as that obtained when the correct signal model is used     

Comments on "Optimum Bandwidth of a Low-Pass Filter for Detection of a Pulse in Nonstationary Noise"

In a recent correspondence1 a calculation of the optimum bandwidth of a low-pass RC filter for the detection of a pulse signal in nonstationary noise was presented. The purpose of this correspondence is: 1) to point out additional references to the work which has been conducted in the stationary noise case, and 2) to present an interesting alternate derivation of the expected output noise power for the nonstationary noise case.     

Synchronization effects on probability of pulse train interception

Expressions for the probability of coincidence between pairs of pulse trains are examined. The application to systems of interest in the field of electronic warfare is emphasized. Theoretical results for the limit of uncorrelated pulse trains are reviewed and new theory is developed to cover the case of coincidence between uniform pulse trains where the coincidence phenomena may be highly correlated. Monte Carlo experiments show that increasing the jitter level on one of the pulse trains causes a transition between the correlated and uncorrelated regimes     

Bayesian Quickest Signal Detection in a Discrete-Time Observation

This paper deals with the problem of quickest detection of a signal in discrete-time observations where the noise is not necessarily additive. By introducing a new cost function, penalizing the decision delay, in addition to penalizing wrong decisions as in the classical case, a global risk function is derived for use in a Bayesian framework. The minimization of the average risk leads to the optimum Bayesian decision regions, giving the structure of the optimum receiver. Some simplifications for elementary costs and some applications are investigated. The optimum receiver is shown to be a parallel bank of classical optimum filters, each one matched to a particular delay of the signal to be detected. Our approach is shown to apply to the detection of certain changes in a stochastic process.