A recursive method for calculating binary integration detectionprobabilities

The calculation of the probability of detection for a binary integration when the probability of a threshold crossing changes from sample to sample is presented. The significance of the algorithm is that it uses a simple recursion relation which provides a computationally efficient means of performing the calculation     

Mean-Level Detection of Nonfluctuating Signals

Analysis of the performance of a mean-level threshold in the detection of nonfluctuating signals is performed. Formulas for the probability of detection are derived and a simple recursive method that can be used for computations is described. Binary integration is discussed, and it is shown that the loss in sensitivity due to the use of an adaptive threshold followed by binary integration is only a fraction of a decibel when compared with optimum binary integration. Binary integration results are given for both fluctuating and nonfluctuating signals.     

An Alternative Method for Analyzing the Double Threshold Detector

An alternative method for analyzing the performance of a double threshold or M-out-of-N detector is discussed. Detection performance for the suggested method is based on the probability that a return crosses the threshold for the Mth time (a detection is declared) on the kth return or look. It is shown that this formulation has many advantages, as compared with the conventional method of analysis which employs the binomial probability distribution, since the upper limit N is not contained in the resulting probability expressions. It is shown that the probability of detection obtained by the alternate method is the same as that obtained if the detection method were analyzed as a Markov chain with M+1 states. Use of the method results in simple expressions for the mean and variance of the number of looks before detection, provides an alternative way of estimating the probability of a threshold crossing, and leads to computationally simple bounds for the probability of false alarm.     

从探测概率的角度评价飞机的隐身性能

采用雷达散射截面(RCS)均值来衡量飞机的隐身性能并不能给出足够充分的信息,从信号检测概率的角度来衡量飞机的可探测性可以提供更完整的信息。本文从探测概率的角度详细分析了4种典型隐身飞机RCS起伏数据与虚警概率、探测概率、探测距离、信噪比(SNR)等参数间的关系,并给出了完整的推导过程,采用此方法可以对任意飞机目标在给定雷达参数下的可探测性进行准确计算。一般在雷达性能评估中需要用到RCS起伏模型,本文对飞机目标回波信号进行检测分析时采用数值计算方法,信号的概率密度函数(PDF)直接来源于原始回波起伏信号,避免了模型拟合带来的回波特性失真,且不会产生大的计算误差。通过对比分析计算结果发现:单脉冲检测下回波信号均值中值比越大,回波取值范围越小,则信号的检测概率越小,其中均值中值比一般相差3倍以上时可以得到明显不同的检测概率;在快起伏假设下,非相参积累检测的积累脉冲数Nin即便较小,也能得到较大的信噪比增益,此增益可能大于Nin。     

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     

Pade approximations of probability density functions

The analysis of radar detection systems often requires extensive knowledge of the special functions of applied mathematics, and their computation. Yet, the moments of the detection random variable are often easily obtained. We demonstrate here how to employ a limited number of exactly specified moments to approximate the probability density and distribution functions of various random variables. The approach is to use the technique of Pade approximations (PA) which creates a pole-zero model of the moment generating function (mgf). This mgf is inverted using residues to obtain the densities     

Frequency-Agile Radar Signal Processing

Modern radars may incorporate pulse-to-pulse carrier frequency modulation to increase probability of detection, to reduce Vulnerability to jamming, and to reduce probability of interception. However, if coherent processing is used for clutter rejection, the frequency of N consecutive pulses must be held constant for N-pulse clutter cancellation or Doppler filtering. If M pulses are transmitted during the time the antenna illuminates a target, there are M/N coherently integrated echoes available for noncoherent integration in the computer or the operator's display to further improve the signal-to-noise ratio (SNR). In this paper, analytical and simulation methods are employed to determine the balance between coherent and noncoherent integration that yields the greatest SNR improvement. Attention is focused upon a model using peak selection of fast Fourier transform (FFT) Doppler channels and is compared to a reference model involving only a single Doppler channel. Curves of detectable SNR as a function of M and N are presented for both models.     

Cyclic code shift keying: a low probability of intercept communication technique

A low probability of intercept (LPI), or low probability of detection (LPD) communication technique known as cyclic code shift keying (CCSK) is described. We discuss the basic concepts of CCSK and describe a system based on the use of random or pseudorandom codes for biphase modulation. We use simulation to show that the bit error rate (BER) for CCSK can be closely estimated by using existing equations that apply to M-ary orthogonal signaling (MOS). Also, we show that significantly fewer computations are required for CCSK than for MOS when the number of bits per symbol is the same. We show that using biphase modulation results in waveforms that have a large time-bandwidth product and very low input signal-to-noise ratio (SNR) and thus inherently have an LPI by a radiometer. We evaluate detection by a radiometer and show that LPI can be achieved by using codes of lengths greater than about 2/sup 12/ (i.e., by transmitting more than about 12 bits per symbol). Results illustrate the effect that the CCSK symbol length and error probability, and the radiometer integration time and probability of false alarm (PFA), have on detection by a radiometer. We describe a variation of CCSK called truncated CCSK (TCCSK). In this system, the code of length 2/sup k/ is cyclically shifted, then truncated and transmitted. Although shortened, the truncated code still represents k bits of information, thus leading to an increased data rate. We evaluate radiometer detection of TCCSK and it is shown that the probability of detection is increased compared with the detection of CCSK.     

Detection Loss of the Sample Matrix Inversion Technique

The sample matrix inversion (SMI) technique is used for Doppler and/or array processing. Previous analysis of the technique has been in terms of signal-to-interference plus noise ratio (SINR). For Gaussian statistics, this performance measure gives the same loss values as does a probability of detection analysis for linear-time invariant systems. It is often somewhat less valid for nonlinear or time variant systems. As SMI is a nonlinear technique, a probability of detection analysis has been performed. It is shown that the detection loss is larger than that computed by the SINR measure. It is also shown that though the loss predicted by the SINR measure only depends upon the number of measurements used to estimate the covariance matrix, the detection loss depends upon the false alarm probability and the number of adaptable elements in addition to the number of measurements.     

Noncoherent Detection of Split-Phase FSK by Multiple Predetection Filtering

When the cumulative drift in the center frequency of a binary split-phase FSK signal exceeds the peak deviation of the signal, a conventional noncoherent receiver (i.e., one provided with only two IF filters) may be unable to achieve the probability of error per bit which the designer desires. This limitation may be overcome if the receiver is provided with a bank of more than two contiguous filters (each followed by an envelope detector) tospan the total IF band the instantaneous IF signal might occupy. It is shown that the probability of error per bit for such a receiver is a function of 1) the ratio F of peak frequency deviation to peak frequency drift, 2) the number M of IF filter/detectors, and 3) the signal-to-noise ratio ? in the output of the filter containing the signal. It is further shown thatfor a given value of F an increase in M reduces the amount of transmitter power the communication system designer must provide to yield a given probability of error per bit.     

Adaptive False Alarmrm Regulation in Double Threshold Radar Detection

In practical situations the false alarm probability in double threshold radar detection, sometimes known as binary integration with sliding window detection, is dependent on the nonstationarity and azimuthal correlation of the clutter which is present. Control of the false alarm probability can be achieved, to a certain extent, by the adjustment of the second threshold in the detection process. In this study two adaptive control techniques which are based on the statistical characteristics of the data are compared. Comparing the results for a technique based on first-order statistics with one based on second-order statistics, it is shown that the second-order, or correlation sensitive, technique can give a reduction of 30 to 45 percent in the false alarm probability with no corresponding loss in the detection probability. An interesting aspect of the results is the fact that the effects of the size of the sample area and the bias in the correlation estimator are clearly evident.     

Blind adaptive decision fusion for distributed detection

We consider the problem of decision fusion in a distributed detection system. In this system, each detector makes a binary decision based on its own observation, and then communicates its binary decision to a fusion center. The objective of the fusion center is to optimally fuse the local decisions in order to minimize the final error probability. To implement such an optimal fusion center, the performance parameters of each detector (i.e., its probabilities of false alarm and missed detection) as well as the a priori probabilities of the hypotheses must be known. However, in practical applications these statistics may be unknown or may vary with time. We develop a recursive algorithm that approximates these unknown values on-line. We then use these approximations to adapt the fusion center. Our algorithm is based on an explicit analytic relation between the unknown probabilities and the joint probabilities of the local decisions. Under the assumption that the local observations are conditionally independent, the estimates given by our algorithm are shown to be asymptotically unbiased and converge to their true values at the rate of O(1/k/sup 1/2/) in the rms error sense, where k is the number of iterations. Simulation results indicate that our algorithm is substantially more reliable than two existing (asymptotically biased) algorithms, and performs at least as well as those algorithms when they work.     

Adaptive Radar Detection with Asymptotically Regulated False-Alarm Rate

An adaptive detection procedure is described by which the detection threshold is so adjusted as to provide an asymptotic false-alarm probability PFA that is approximately invariant with changes in radar clutter return amplitude probability density functions (pdf's) in a broad class. The class includes Rayleigh, chi, Weibull, and lognormal pdf's. The receiver noise is also taken into account. The clutter-plus-noise pdf is approximated by a truncated generalized Laguerre series, the coefficients of which are estimated from the radar returns using "cell averaging" techniques. This estimation is assumed to be perfect. The results obtained indicate that the "bias" error, defined as the normalized difference between the design PFA and the asymptotic PFA corresponding to the computed threshold, lies within a fraction of an order of magnitude for 10-3?PFA ? 10-6. For PFA ?10-6 the bias error is more than an order of magnitude. These results are for the case when a single independent radar return is processed at a time. The bias error decreases as the number of postdetection integrations of independent returns increases.     

Spatially distributed target detection in non-Gaussian clutter

Two detection schemes for the detection of a spatially distributed, Doppler-shifted target in non-Gaussian clutter are developed. The non-Gaussian clutter is modeled as a spherically invariant random vector (SIRV) distribution. For the first detector, called the non-scatterer density dependent generalized likelihood ratio test (NSDD-GLRT), the detector takes the form of a sum of logarithms of identical functions of data from each individual range cell. It is shown under the clutter only hypothesis, that the detection statistic has the chi-square distribution so that the detector threshold is easily calculated for a given probability of false alarm P_F. The detection probability P_D is shown to be only a function of the signal-to-clutter power ratio (S/C)_opt of the matched filter, the number of pulses N, the number of target range resolution cells J, the spikiness of the clutter determined by a parameter of an assumed underlying mixing distribution, and P_F. For representative examples, it is shown that as N, J, or the clutter spikiness increases, detection performance improves. A second detector is developed which incorporates a priori knowledge of the spatial scatterer density. This detector is called the scatterer density dependent GLRT (SDD-GLRT) and is shown for a representative case to improve significantly the detection performance of a sparsely distributed target relative to the performance of the NSDD-GLRT and to be robust for a moderate mismatch of the expected number of scatterers. For both the NSDD-GLRT and SDD-GLRT, the detectors have the constant false-alarm rate (CFAR) property that P_F is independent of the underlying mixing distribution of the clutter, the clutter covariance matrix, and the steering vector of the desired signal     

Calculation of Probability of Detection with Target Scintillation

This paper presents a computational method for the calculation of probability of detection using measured radar target cross-section data. The described method can also be used for probability of detection calculations when the radar target cross section follows a specified probability density function. Using the computational procedure of the paper, a number of curves are generated which can be used for probability of detection calculations with exponential and Gaussian radar target cross-section distributions. The results obtained using theoretical distributions are compared with the corresponding results using actual target cross-section measurements. The results of computer runs are compared to the corresponding values in the literature where available.     

Cascaded detector for multiple high-PRF pulse Doppler radars

A postdetection design methodology for a multiple high-pulse-repetition frequency (PRF) pulse Doppler radar has been developed. The postdetection processor consists of an M out of N detector where range and target ambiguities are resolved, followed by a square-law detector which enhances the minimum signal-to-noise (S/N) power-ratio per pulse burst performance. For given probabilities of false alarm and detection, formulas are derived from which the three thresholds associated with the cascaded detector can be found. Fundamental tradeoffs between the minimum S/N required, number of ghosts, and the number of operations (NOPs) that the cascaded detector must perform are identified. It is shown that the NOPs and the number of ghosts increase and the minimum S/N required decreases as the binary M out of N detector passes more detections to the square-law detector     

A Method for Setting a Test Signal Level Relative to Noise

A method is described for adjusting the leval of an RF test signal generator relative to the noise level at the receiver output. The method compares a detected output to a threshold and counts the number of times noise and signal plus noise cross the threshold in a given number of tries. By setting the threshold at a given false alarm probability for noise alone and then adding the test signal and adjusting its level to give a specified detection probability, the signal-to-noise ratio can be calibrated to an accuracy that depends on the number of samples used to measure the probabilities. The false alarm and detection probabilities are given for best accuracy as well as the rms error in signal-to-noise ratio as a function of the number of samples used.     

裂纹检测概率曲线的统计测定

本文提出了一种估计裂纹检测概率曲线的试验方法和数据处理方法。给出的估计检测概率置信下限的公式精确、简单,并附有实例。文中还提出了估计裂纹检测概率曲线的工程简化方法。     

A Useful Approximation to Binary FSK Probability of Error

A technique for calculating a simple approximation to the probability of error for binary frequency-shift keying (FSK) systems when the predetection bandwidth-bit duration product is large enough so that the signal portion of the detector output is relatively undistorted is described. The results are mathematically simple and avoid the calculation of the density function of the nonclick noise part of the output phase noise.     

Asymptotic Performance of Sparse Signal Detection Using Convex Programming Method

The detection of sparse signals against background noise is considered. Detecting signals of such kind is difficult since only a small portion of the signal carries information. Prior knowledge is usually assumed to ease detection. In this paper, we consider the general unknown and arbitrary sparse signal detection problem when no prior knowledge is available. Under a Neyman-Pearson hypothesis-testing framework, a new detection scheme is proposed by combining a generalized likelihood ratio test (GLRT)-like test statistic and convex programming methods which directly exploit sparsity in an underdetermined system of linear equations. We characterize large sample behavior of the proposed method by analyzing its asymptotic performance. Specifically, we give the condition for the Chernoff-consistent detection which shows that the proposed method is very sensitive to the 2 norm energy of the sparse signals. Both the false alarm rate and the miss rate tend to zero at vanishing signal-to-noise ratio (SNR), as long as the signal energy grows at least logarithmically with the problem dimension. Next we give a large deviation analysis to characterize the error exponent for the Neyman-Pearson detection. We derive the oracle error exponent assuming signal knowledge. Then we explicitly derive the error exponent of the proposed scheme and compare it with the oracle exponent. We complement our study with numerical experiments, showing that the proposed method performs in the vicinity of the likelihood ratio test (LRT) method in the finite sample scenario and the error probability degrades exponentially with the number of observations.