IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 1, JANUARY 2007
21
袁泉歌曲On the Energy Detection of Unknown Signals Over Fading Channels
肉末炖豆腐Fadel F. Digham, Member, IEEE, Mohamed-Slim Alouini, Senior Member, IEEE, and Marvin K. Simon, Fellow, IEEE
Abstract—This letter address the problem of energy detection of an unknown signal over a multipath channel. It starts with the no-diversity ca, and prents some alternative clod-form expressions for the probability of detection to tho recently reported in the literature. Detection capability is boosted by implementing both square-law combining and square-law lection diversity schemes. Index Terms—Diversity schemes, energy detection, fading channels, low-power applications, square-law detector, unknown signal detection.
numerical examples are demonstrated in Section VI, and concluding remarks are offered in Section VII. II. SYSTEM MODEL AND NOTATIONS While our analysis applies to either low-pass (LP) or bandpass (BP) systems, we here focus on the BP reprentation. The received BP waveform can be
reprented as (1) where denotes the real part operation, is a slowfading channel with amplitude and pha , is the carrier and refer to the two hypothes of signal frequency, both prence and signal abnce, respectively, is an equivalent LP reprentation of the unknown signal and denoting the in-pha (I) and quadrature with (Q) components, respectively, and likewi, is an equivalent LP AWGN process with a zero mean and a known flat power spectral density (PSD). If denotes the the signal bandwidth (i.e., positive one-sided noi PSD and . spectrum support), the noi variance will, in turn, be Also, I and Q components will be each confined to the frequency and the one-sided noi PSD of either support or will be (to rerve a noi variance of ). shall denote the signal-to-noi ratio (SNR) is the signal energy. where The receiver structure can generally be described as follows. The received signal is first prefiltered by an ideal BP filter. Then, the output of this filter is squared and integrated over a time to finally produce a measure of the energy of the interval received waveform. The output of the integrator denoted by acts as a test statistic to test the two hypothes and . , assuming a narrowband signal, and relying on the Under sampling theorem approximation1 [2], can be expresd as
I. INTRODUCTION HE PROBLEM of detecting an unknown deterministic signal over a flat bandlimited Gaussian noi channel was first addresd by Urkowitz [10]. In his proposal, the receiv
er consisted of an energy detector which measures the energy in the received waveform over an obrvation time window. This energy-detection problem has been revisited recently by Kostylev in [5] for signals operating over a variety of fading channels. Our contribution in this letter is twofold. First, we prent an alternative analytical approach to the one prented in [5] and obtain clod-form expressions for the probability of detection over Rayleigh and Nakagami fading channels. Second, and more importantly, we quantify the improvement in detection capability (specially for relatively low-power applications) when low-complexity diversity schemes such as square-law combining (SLC) and square-law lection (SLS) are implemented. While diversity analysis is carried out for independent Rayleigh channels for the SLS scheme, both independent and correlated cas are considered for the SLC one. For more details, the reader is referred to [2, Ch. 9]. The rest of this letter is organized as follows. The system model is described in Section II. Conditional (on the fading) probability of detection , and probability of a fal alarm , are evaluated in Section III over additive white Gaussian noi (AWGN) channels. While Section IV deduces the probabilities over Rayleigh and Nakagami fading channels, Section V studies the impact of diversity for Rayleigh channels. Finally,
T
调查报告封面Paper approved by F. Santucci, the Editor for Transmission Systems of the IEEE Communications Society. Manuscript received December 9, 2003; revid March 30, 2005 and April 23, 2006. This work was supported by the ARL Communications and Networks CTA, under Cooperative Agreement DAAD19-01-20011. This paper was prented in part at the IEEE International Conference on Communications, Anchorage, AK, May 2003. F. F. Digham is with the Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN 55455 USA (e-mail: fdigham@ece. umn.edu). M.-S. Alouini was with the Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN 55455 USA. He is now with the Department of Electrical Engineering, Texas A&M University at Qatar, Doha, Qatar (e-mail: alouini@qatar.tamu.edu). M. K. Simon is with the Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109-8099 USA (e-mail: marvin.k.simon@jpl. v). Digital Object Identifier 10.1109/TCOMM.2006.887483
(2) where is the number of samples per either I or Q comterms to sum over), , ponents (we finally have , and generally and , respectively, denote the th samples of and , i.e., I and Q components. It then follows that under , has a noncentral chi-square distribution
1If a signal is assumed to be bandlimited and in the same time obrved over a limited interval T , th
e sampling theorem will then yield an approximate expression of the signal in terms of its limited samples.
0090-6778/$25.00 © 2007 IEEE
22
IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 1, JANUARY 2007
伊丽莎白一世with variance , noncentrality parameter2 , and degrees of freedom (DOFs)3 . Likewi, given will be central chi-square distributed. In the rest of the letter, we shall and (for some positive u the generic parameters value ). The results can then be specialized to the prented and . Now, the probability density ca study for function (PDF) of can be expresd as
where is the Nakagami parameter and is the average SNR. for a Nakagami With the aid of Appendix A, the average , can be obtained as channel,
(7) where hypergeometric function teger , we have , is the confluent [3, Sec. 9.2], and for in-
(3)
where is the gamma function [3, Sec. 8.31] and is the th-order modified Besl function of the first kind [3, Sec. 8.43]. III. DETECTION AND FALSE ALARM PROBABILITIES OVER AWGN CHANNELS An approximate expression for the probability of detection over AWGN channels was prented in [10]. In this cand tion, we prent exact clod-form expressions for both the probability of a fal alarm , which are defined as and , respectively, where is a decision threshold. Bad on the statistics of , can be evaluated as (4) (9) is the incomplete gamma function [3]. This result where matches the one obtained in [5, eq. (19)] for the aforementioned . Making u of [9, eq. (2.1-124)], the cumulaca with tive distribution function (CDF) of can be obtained (for even ), and ud to evaluate as (5) where is the generalized Marcum -function [8]. (8) where is the Laguerre polynomial of degree [3, eq. (8.970)]. As a by-product of the above result, the average over a Rayleigh channel, , can be obtained by tting in (7). Alternatively, starting with Rayleigh distribution [tting in (6)] can be obtained as4
V. DETECTION AND FALSE ALARM PROBABILITIES WITH DIVERSITY RECEPTION In this ction, we look into the energy-detection performance when SLC and SLS diversity schemes are employed. While SLC is studied for independent and identically distributed (i.i.d.) as well as correlated Rayleigh fading channels, SLS is studied for the independent ca only. A. Square-Law C
ombining (SLC) In this scheme, the outputs of the square-law devices (squareand-integrate operation per branch), denoted as where is the number of diversity branches, are combined to yield a new decision statistic . Under and for i.i.d. central chi-square variates, AWGN channels, adding each with DOFs and variance , will result in another chisquare variate with DOFs and the same variance . Therefore, analogous to (4), we have (10)
4This expression in [1, eq. (16)] (with = 1, a = 2, and N=2 u) has a typo (the last exponent i was missing). An alternative expression to that in (9) was also obtained in [4].
IV. PROBABILITY OF DETECTION OVER FADING CHANNELS WITH NO DIVERSITY In this ction, we derive the average detection probability in over Nakagami and Rayleigh fading channels. Notice that (4) in this ca will remain the same, since it is independent of the SNR. The PDF of over a Nakagami channel is given by (6)
2The noncentrality parameter can be obtained from (2) after tting n = n = 0 and using the sampling theory approximation E ' (1=2) 1 (1=W ) (S + S ), with the (1=2) factor reflecting the energy mapping from equivalent LP to BP signals. 3
sample position (at 0 or 0 ).
论语八佾篇
N can be approximated as either 2(TW) or 2(TW+1), depending on the first
IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 1, JANUARY 2007
JITTER23
Likewi, under , will be a chi-square variate with DOFs, noncentrality parameter , and variance . Hence, at the combiner output for AWGN channels, , can be evaluated by analogy to (5) as (11) In the following, this probability of detection is averaged over both i.i.d. and correlated Rayleigh channels. 1) i.i.d. Rayleigh Channels: The PDF of for i.i.d. Rayleigh branches is quite similar to that in (6), while replacing by and each by . Hence, is equiveach in (7), after replacing each by , each by alent to , and each by . 2) Correlated Rayleigh Channels: For correlated Rayleigh branches, the PDF of is given by [6, eq. (10-60)]
Fig. 1. Complementary ROC curves for a Rayleigh channel.
(12) Averaging this and using over independent Rayleigh branches, in (9) yields
, ’s are the eigenvalues of the covariance matrix , , with denoting the power-correlation coefficient between and . Using the PDF expression in (12), which reprents a weighted in (9) sum of exponen
tial variates, and expressing , the probability of detection in (11) can be as averaged, yielding where , and (13) For the special ca when , , and , it follows that and [6]. In addition, the ca of independent, but not identically distributed, branches can be deduced by simply having a diagonal with being the th diagonal entry. B. Square-Law Selection (SLS) In the SLS diversity scheme, the branch with maximum decision statistic is to be lected [7]. variates, for SLS, , Under , and given i.i.d. given , , can be evaluated using the CDF of yielding
(16)
VI. NUMERICAL EXAMPLES We quantify the receiver performance by depicting the receiver operating characteristic (ROC) ( versus ), or equiva, lently, complementary ROC (probability of a miss versus ) for different situations of interest. In the following examples, and , as is the ca from (2). Fig. 1 illustrates the complementary ROC over a Rayleigh channel. The right part of this figure asrts the fact that for the same signal energy, the fewer the samples, the better the performance, as is increas for a given . The left part of this the ca when figure shows another scenario, in which the signal is time-limited over a , interval and can be approximated as a bandlimited one. In particular, we study the effect of choosing when is unknown. Considering a raid cosine signal , implies it is shown here that overdetermining . This better performance than underdetermining it ass
erts the expectation of higher performance nsitivity to a loss in signal energy, rather than to an increa in noi energy. Fig. 2 quantifies the performance gain as the Nakagami param20 dB. For example, there is roughly a eter increas for perspective for gain of one order of magnitude from the , compared with the Rayleigh ca .
(14) Similarly, and conditioning on (under ), AWGN channels, , can be obtained as for SLS over回家用英语怎么说
(15)
24
IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 55, NO. 1, JANUARY 2007
APPENDIX A EVALUATION OF
IN
(7)
Averaging (5) over (6) while using the change of variable yields (17)
,nessie
, , , and . can be recursively evaluated with the aid For of [8, eq. (29)], yielding where ,
Fig. 2. Complementary ROC curves for a Nakagami channel ( = 20 dB, N = 10).
. . . (18)
where
(19)
(20)
and
can be evaluated with the aid of [8, eq. (25)] for integer (i.e., integer ), and given that the first-order Marcum -function . REFERENCES
[1] F. F. Digham, M.-S. Alouini, and M. K. Simon, “On the energy detection of unknown signals over fading channels,” in Proc. IEEE Int. Conf. Commun., Anchorage, AK, May 2003, pp. 3575–3579. [2] F. F. Digham, “On signal transmission and detection over fading channels,” Ph.D. disrtation, Univ.
Minnesota, Minneapolis, MN, Jul. 2005. [3] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, 6th ed. San Diego, CA: Academic, 2000. [4] V. I. Kostylev, “Characteristics of energy detection of quasideterministic radio signals,” in Proc. Radiophys. Quantum Electron., Oct. 2000, vol. 43, pp. 833–839. [5] ——, “Energy detection of a signal with random amplitude,” in Proc. IEEE Int. Conf. Commun., New York, NY, May 2002, pp. 1606–1610. [6] W. C. Y. Lee, Mobile Communications Engineering. New York: McGraw-Hill, 1982. [7] E. A. Neasmith and N. C. Beaulieu, “New results on lection diversity,” IEEE Trans. Commun., vol. 46, no. 5, pp. 695–704, May 1998. [8] A. H. Nuttall, “Some integrals involving the Q -function,” Naval Underwater Syst. Center (NUSC) Tech. Rep., May 1974. [9] J. G. Proakis, Digital Communications, 4th ed. New York: McGrawHill, 2001. [10] H. Urkowitz, “Energy detection of unknown deterministic signals,” Proc. IEEE, vol. 55, no. 4, pp. 523–531, Apr. 1967.
Fig. 3. Complementary ROC curves for dual-branch diversity systems over Rayleigh channels ( = 20 dB, N = 10).
Finally, the effect of diversity over Rayleigh branches is il20 dB and . SLC and SLS lustrated in Fig. 3 for schemes with i.i.d. branches provide mostly the same gain of perspective, comat least one order of magnitude from the pared with the no-diversity ca. Approximately half of this gain is lost when e
mploying the SLC scheme with two correlated . branches of VII. CONCLUSION Using a sampling theory-bad approach, we studied the performance of an energy detector for an unknown transmit signal under both AWGN and fading channels. We also quantified the improvement in detection capability when receive diversity schemes are employed. The analysis of this energy detector is timely for emerging applications involving ultra-wideband and cognitive radio technologies.
