0% found this document useful (0 votes)
34 views5 pages

Maximum-Likelihood Magnitude-Based Amplitude and Noise Variance Estimation

Uploaded by

saadwaseem786
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
34 views5 pages

Maximum-Likelihood Magnitude-Based Amplitude and Noise Variance Estimation

Uploaded by

saadwaseem786
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

414 IEEE SIGNAL PROCESSING LETTERS, VOL.

28, 2021

Maximum-Likelihood, Magnitude-Based, Amplitude


and Noise Variance Estimation
Ming-Wei Wu , Member, IEEE, Yan Jin, Yan Li , Tianyu Song , and Pooi-Yuen Kam , Fellow, IEEE

Abstract—Maximum likelihood (ML) amplitude and noise vari- and covariance E[n(k)n∗ (k − j)] = 2σ 2 δ(j), where δ is the
ance estimation without having to jointly estimate the frequency Kronecker delta. It is assumed that there is no prior knowledge
and the phase and based only on information from the noisy re- on the unknown nonrandom parameters A, ω and θ, i.e., pa-
ceived signal magnitude, is studied for a single sinusoid in complex rameter A can be any positive, real number, and the modulo
additive white Gaussian noise. This estimation problem is equiva-
lent to the classic problem of parameter estimation for the Rician
−2π reduced ω and the modulo −2π reduced θ can take any
distribution. While solving the likelihood equation is impossible in value over the interval [−π, π). With these assumptions, the
general, we propose a new approach based on a large argument most common estimation approach is to apply the theory of
approximation. For the case with known noise variance, a closed- maximum likelihood (ML) for joint amplitude, frequency and
form ML amplitude estimator is obtained, which outperforms phase estimation. However, due to the nonlinear nature of the
the conventional root-mean-square estimator. For the case with problem, there is so far no explicit, closed-form solution for the
unknown noise variance, the closed-form joint amplitude and noise absolute maximum of the likelihood function. This has led to
variance estimators obtained do not require prior knowledge of one the fast-Fourier-transform-based frequency domain solution [1]
another. which, through the sequential computation of [1, eqs. (29), (30)
Index Terms—Amplitude estimation, cramer-rao bound, and (28)], results in the estimates of ω, θ and A, in the order
maximum likelihood (ML) estimation, noise variance estimation, as mentioned. Hence, the optimal ML estimate  can only be
rician distribution. obtained after obtaining the optimal ML estimates ω̂ and θ̂. It
is also known that there are some obstacles in estimating ω
and θ, such as outliers, computational complexity and phase
I. INTRODUCTION unwrapping, etc [1]–[3]. Therefore, it would be desirable to find
STIMATING the unknown parameters A, ω and θ from the an estimator  that is independent of ω̂ and θ̂ for convenience.
E noisy, discrete-time, complex data samples r(k), where This can be achieved by making use of only the magnitude
information of the data samples, i.e. {|r(k)|}k .
r(k) = Aej(ωk+θ) + n(k), k = 0, 1, 2, . . . (1) The magnitude-based amplitude estimation problem is equiv-
is an important problem in communication engineering. The alent to the classic problem of parameter estimation for the
additive, white Gaussian noise (AWGN) {n(k)} is a sequence Rician distribution [4]. For example, in otoacoustic emission
of circularly symmetric, independent and identically distributed (OAE) measurement [5] with signal model r(k) = Aejθ +
(i.i.d.), complex, Gaussian random variables with mean zero n(k), the measurement noise n(k) is found to be reasonably
well approximated by a circularly symmetric Gaussian noise,
Manuscript received August 17, 2020; revised January 5, 2021; accepted
but the sound pressures are typically reported as magnitudes
January 13, 2021. Date of publication January 29, 2021; date of current version without knowledge of the phase, i.e. {|r(k)|}k [5]. Estimation
March 4, 2021. The work was supported by the National Natural Science of magnitude A using {|r(k)|}k is equivalent to estimation of
Foundation of China under Grant 61571316. The associate editor coordinating the line-of-sight component of the Rician distribution. Similarly,
the review of this manuscript and approving it for publication was Dr. Luis the problem lies in areas like magnetic resonance imaging [6]–
Antonio Azpicueta-Ruiz. (Corresponding author: Yan Li.)
Ming-Wei Wu is with the School of Information and Electronic Engineering, [9], speech processing [10], synthetic aperture radar [11], [12],
Zhejiang University of Science and Technology, Hangzhou 310023, China (e- sonar [13] and channel modeling [14], [15].
mail: [email protected]). ML amplitude estimation is proposed in [7]. However, due
Yan Jin is with the School of Information and Electronic Engineering, Zhe- to the complicated form of the derived ML condition, existing
jiang University of Science and Technology, Hangzhou 310023, China, and also
with the Center for Advanced Microelectronic Devices, NUS (Suzhou) Research
methods are not able to find explicit solutions. Therefore, numer-
Institute (NUSRI), Suzhou 215123, China (e-mail: [email protected]). ical methods are resorted to, such as the iterative Brent’s algo-
Yan Li is with the Department of Electrical and Computer Engineering rithm in [7] and search using numerical optimization in [8], both
(ECE), National University of Singapore (NUS), Singapore 117583 (e-mail: of which have high computational complexity. In addition, [6]–
[email protected]). [8] all require prior knowledge of the noise variance parameter
Tianyu Song was with the Department of Electrical and Computer Engineer-
ing (ECE), National University of Singapore (NUS), Singapore 117583, and σ 2 . For noise variance estimation, [16] proposed a magnitude-
also with the Center for Advanced Microelectronic Devices, NUS (Suzhou) based estimator using the method of moments with quadratic
Research Institute, Suzhou 215123, China. He is now with Huawei Device, sample mean and fourth-power sample mean, but it does not
Shenzhen 518129, China (e-mail: [email protected]). attempt to optimize the performance. A joint ML amplitude and
Pooi-Yuen Kam is with the School of Science and Engineering, Chinese
University of Hong Kong, Shenzhen 518172, China, and also with the Center for
noise variance estimator is proposed in [8] by maximizing the
Advanced Microelectronic Devices, NUS Research Institute (NUSRI), Suzhou two-dimensional log-likelihood function of A and σ 2 . Again,
215123, China (e-mail: [email protected]). its optimization cannot be solved analytically and numerical
Digital Object Identifier 10.1109/LSP.2021.3055464 optimization with high complexity must be resorted to.
1070-9908 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: National University Fast. Downloaded on October 31,2024 at 08:48:55 UTC from IEEE Xplore. Restrictions apply.
WU et al.: MAXIMUM-LIKELIHOOD, MAGNITUDE-BASED, AMPLITUDE AND NOISE VARIANCE ESTIMATION 415

Remark: Although the Rician PDF (4) can be obtained directly


from (1) by noting that r(k) = [A cos(ωk + θ) + [n(k)]] +
j[A sin(ωk + θ) + [n(k)]], where [n(k)] and [n(k)] are the
inphase and quadrature component of n(k) in the I-Q coordinate,
respectively, we use Fig. 1 and (2) to derive (4). The reason is
two-fold:
1) To illustrate geometrically that the amplitude estimator [1,
eq. (24)] is obtained from the measurements A + nI (k) which
is real part of the product of r(k) and e−j(ωk+θ) (note that
Fig. 1. Geometric representation of the received signal. the quantity e−j(ωk+θ) can be absorbed into the AWGN n(k)
without changing its statistical properties). Needless to say, for
In this paper, we adopt an approximate approach to solve the case of unknown ω and unknown θ, e−j(ωk+θ) can only be
the ML estimation problem explicitly. Using a large argu- computed from the estimates ω̂ and θ̂.
ment approximation of the ML condition, we obtain closed- 2) To show geometrically the relationship between |r(k)| and
form estimators that are easy to implement, for both the cases A + nI (k). It is seen from Fig. 1 and (2) that if the quantity
of known and unknown noise variance. Section II presents nQ (k) is small, A + nI (k) can be approximated by |r(k)|.
the signal model and conventional amplitude estimators. In Estimating A from the measurements A + nI (k) is equivalent
Section III, a closed-form ML amplitude estimator with known to finding the mean of a Gaussian PDF, and it is well-known that
noise variance is derived, which outperform the conventional this is given by the sample mean of the measurements, defined
root-mean-square (RMS) estimator. In Section IV, it is first in Section II-B.
proven that the quadratic sample mean is the ML estimate of the
second absolute moment. Then, we obtain the closed-form joint B. Conventional Sample Mean and RMS Estimators
amplitude and noise variance estimators, which only require
the magnitude of the received signal and do not require prior The performance of an estimator  with respect to the un-
knowledge of one another. Conclusions are made in Section V. known constant parameter A is measured by the mean square
error (MSE). As the estimator  in general can be complex, the
II. SIGNAL MODEL AND CONVENTIONAL ESTIMATORS MSE can be expressed as

A. Signal Model E[|Â − A|2 ] = E[((Â) − A)2 ] + E[2 (Â)]. (5)


Fig. 1 gives a geometric phasor representation of Aej(ωk+θ) , Since the values of r(k) are complex numbers, there are
n(k) and r(k) of (1) in the inphase-and-quadrature (I-Q) co- two different sample mean estimators, using the real part or
ordinate complex plane. The complex AWGN n(k) is decom- magnitude of r(k), depending on the knowledge of the received
posed into two orthogonal components, nI (k) and nQ (k), with r(k) at the receiver:
nI (k) being parallel to, and nQ (k) perpendicular to the phasor 1) When the complex value of r(k) can be obtained accu-
Aej(ωk+θ) . Since n(k) is circularly symmetric, the two orthog- rately, i.e. by coherent detection, by taking the real part of r(k)
onal components nI (k) and nQ (k) are statistically i.i.d., real- −1
only, i.e. ÂSMr = N1 N k=0 [r(k)], where N is the number of
valued, Gaussian random variables with mean zero and variance data sample measurements, the imaginary part is eliminated,
σ 2 . The signal-to-noise ratio (SNR) is defined as γ = A2 /2σ 2 . reducing the noise power by half. It is the optimum receiver
It is seen from Fig. 1 that in the newly formulated coordinate with its MSE equal to the the CRLB given by CRLBr = σ 2 /N
system, which is the original I-Q-coordinate system rotated [1, eq. (18)]. ÂSMr requires coherent reception and, therefore,
counterclockwisely by the angle ωk + θ, the quantity r(k) in is not applicable to the scenario considered in this paper. Its
(1) can be rewritten as performance is only used as a theoretical benchmark.
r(k) = A + nI (k) + jnQ (k). (2) 2) When the complex value of r(k) is not known but only
the magnitude |r(k)| is known, the magnitude sample mean
In the newly formulated coordinate system, the magnitude of estimator is given by
r(k) can be obtained as
 N −1
1
|r(k)| = [A + nI (k)]2 + [nQ (k)]2 . (3) ÂSMm = |r(k)|. (6)
N
k=0
It follows from [4, eqs. (2.1–140) and (2.1–141)] that |r(k)|2 has Both sample mean estimators do not require the knowledge of
a noncentral Chi-square probability density function (PDF) with the AWGN variance σ 2 .
two degrees of freedom and the noncentrality parameter A2 , and The quadratic sample mean of N samples of r(k) is
|r(k)| is a Rician-distributed random variable. The conditional
PDF of |r(k)| is given by 1
N −1
   Â2 M = |r(k)|2 . (7)
p |r(k)| A, σ 2 N
k=0
  
|r(k)| |r(k)|2 + A2 |r(k)|A It is obvious that E[Â2 M ] = E[|r(k)|2 ]. On the other hand,
= exp − I0 (4)
σ2 2σ 2 σ2 the second absolute moment of any complex random variable
where I0 (·) is the modified Bessel function of the first kind of satisfies E[|r(k)|2 ] = A2 + 2σ 2 [17, eq. (5–61)], and hence,
order zero. The goal is to recover A accurately from {|r(k)|}k E[Â2 M ] = A2 + 2σ 2 . By approximating the ensemble mean
with likelihood function given by (4). E[Â2 M ] by the sample mean Â2 M , assuming a large sample
Authorized licensed use limited to: National University Fast. Downloaded on October 31,2024 at 08:48:55 UTC from IEEE Xplore. Restrictions apply.
416 IEEE SIGNAL PROCESSING LETTERS, VOL. 28, 2021

size N , we obtain Â2 M ≈ A2 + 2σ 2 . Assuming that the noise


variance 2σ 2 is known, this leads to the conventional RMS am-
plitude estimator ÂRMS = Â2 M − 2σ 2 . When Â2 M < 2σ 2 ,
which occurs more frequently at low SNR, it is obvious that
ÂRMS results in an imaginary value. Now that A is known to be
real, the imaginary part of estimates is entirely estimation error,
as can be seen in (5). Therefore, we propose to modify the RMS
estimator by taking the real part, i.e.
 
ÂRMSr =  Â2 M − 2σ 2 . (8)

If ÂRMS returns an imaginary value, the estimate ÂRMSr is set to


zero, eliminating the imaginary part and the corresponding MSE.
Similarly, for all subsequent estimators with potential complex
values, we propose to take their real parts.
Fig. 2. MSE of estimator ÂLAr in (16) with known σ 2 , with N = 10, 30.
III. MAXIMUM LIKELIHOOD AMPLITUDE ESTIMATORS WITH
KNOWN NOISE VARIANCE 
ex 3
Assume N measurements of {|r(k)|}N −1 I1 (x) ≈ √ 1− . (12)
k=0 are collected 2πx 8x
−1
for amplitude estimation. Since {nI (k), nQ (k)}Nk=0 is a se-
quence of a pair of i.i.d. Gaussian random variables, we Thus,
have p({|r(k)|}N −1 2 N −1  2
k=0 |A, σ ) = k=0 p(|r(k)| A, σ ). Substi-
tuting (4), the corresponding log-likelihood function is given I1 (x) 4 1
≈1− ≈1− . (13)
by I0 (x) 8x + 1 2x
Λ(A, σ 2 ) = ln p({|r(k)|}N −1 2
k=0 |A, σ ) Using the approximation (13) and after simplications, equation
N −1    (11) can be approximated as
|r(k)| |r(k)|2 + A2 |r(k)|A
= ln − + ln I0 .
σ2 2σ 2 σ2 σ2
k=0 A2 − ÂSMm A + = 0. (14)
(9) 2
Assume that the receiver either has prior knowledge of the noise When the noise variance σ 2 is known at the receiver, it is obvious
variance σ 2 , or has obtained the knowledge via an independent that the two solutions to the quadratic equation (14) are given
estimation method. We want to maximize the right-hand side of by
(9) with respect to A. The necessary condition for  to be the 
ML estimate of A is that it satisfies 1 1
 ÂLA = ÂSMm ± Â2SMm − 2σ 2 . (15)
N −1 2 2
∂Λ(A, σ 2 ) ∂ |r(k)|A NA
= ln I0 2
− 2 = 0. (10)
∂A ∂A σ σ For large argument approximation where Â2SMm  2σ 2 , the
k=0
solution with the − sign results in an estimate approaching
dx I0 (x) =
d
Equivalently, with further simplification using zero. The solution with the + sign is close to ÂSMm and is more
I1 (x)[18, eq. (8.486-9)], we obtain realistic. Simulations show that, in the case of large noise, rare
 
N −1 I |r(k)|A occasions with Â2SMm − 2σ 2 < 0 occur, resulting in complex
1 1 σ 2
A= |r(k)|  . (11) solutions to the quadratic equation in (15). Hence, we propose
N I |r(k)|A to take the real part of (15), i.e.
k=0 02 σ
 
Although the calculation of  by (10) or (11) is conceptually 1 1
ÂLAr = ÂSMm +  Â2SMm − 2σ 2 . (16)
straightforward, it is computationally difficult due to the non- 2 2
linearity in the two special functions I0 (x) and I1 (x). Iterative
algorithms can be resorted to, however, with high computational For lack of space, only performance results obtained via simu-
complexity [19]. We here take the approximate approach to lations are presented. For each SNR value, A = 1 is used and the
solve (10) or (11) by finding some mathematically tractable MSE performance is approximated by the square error averaged
representations of I0 (x) and I1 (x) so that (10) or (11) can be over 106 estimates. Fig. 2 shows that the performance of ÂSMm
solved directly. This can be achieved by approximating I0 (x) improves with SNR and approaches the CRLBr , and hence that
and I1 (x) using only elementary functions. of the optimum receiver ÂSMr . As ÂRMSr uses knowledge of
For large values of x, infinite-term asymptotic approximations σ 2 , it outperforms ÂSMm which uses no knowledge of σ 2 , as
to I0 (x) and I1 (x) are given, respectively, by [20, eq. (9.7.1)]
expected. Although ÂLAr was derived using the large argument
and can be approximated by
 approximation, it outperforms the conventional RMS estimator
ex 1 ÂRMSr and the iterative estimator ÂItr in [19] at all SNR values,
I0 (x) ≈ √ 1+ ,
2πx 8x and has low computational complexity.
Authorized licensed use limited to: National University Fast. Downloaded on October 31,2024 at 08:48:55 UTC from IEEE Xplore. Restrictions apply.
WU et al.: MAXIMUM-LIKELIHOOD, MAGNITUDE-BASED, AMPLITUDE AND NOISE VARIANCE ESTIMATION 417

IV. JOINT MAXIMUM LIKELIHOOD AMPLITUDE AND NOISE


VARIANCE ESTIMATORS
The MSE performance of the above-mentioned estimators are
obtained with perfect knowledge of the noise variance σ 2 . This
may not be practical, as the noise variance is usually estimated
in the first stage before its knowledge is used in amplitude es-
timation. In addition, noise variance estimation is accompanied
by an inevitable estimation error, which usually results in a
performance degradation, as shown in Fig. 2. Therefore we now
look into the more realistic scenario where the noise variance
σ 2 is unknown at the receiver, and we have to estimate it jointly
with the amplitude. Performing ML estimation using the same
approach as in (10), the ML estimate of σ 2 satisfies a
N −1
∂Λ(A, σ 2 ) N |r(k)|2 N A2
= − 2+ k=0
+
∂σ 2 σ 2
2(σ ) 2 2(σ 2 )2 Fig. 3. MSE of estimator Âunr in (21) with unknown σ 2 , with N = 10, 30.
 
N −1 I1 |r(k)|A
σ2 A
− |r(k)|   = 0. (17)
|r(k)|A (σ 2 )2
k=0 I 0 2 σ

We have to solve (11) and (17) jointly. Substituting (11) into


(17) and after simplifications, we arrive at

Â2 M = A2 + 2σ 2 . (18)

Note that Â2 M is the quadratic sample mean defined in (7). We


have shown here in (18) that it is the unbiased ML estimate of
the second absolute moment E[|r(k)|2 ] = A2 + 2σ 2 , in the case
when both A2 and σ 2 are unknown.
We can solve (18) and the large argument approximation (14)
together and obtain the following ML estimate of amplitude and
noise variance as Fig. 4. MSE of estimator σr2 in (22) with unknown A, with N = 10, 30.

2 1
Âun = ÂSMm ± 4Â2SMm − 3Â2 M , (19) their sample averages [16, eq. (13)], i.e.
3 3 
 1 1  14 
2 = 2 Â2 M − 4 Â2 ∓ 2 ÂSMm 4Â2 − 3Â2 M , (20)
σ σ
2 = Â
2M −  2Â 2
− Â4M , (23)
3 9 SMm 9 SMm mr
2 2 2M

where
where ÂSMm is given in (6) and Â2 M is given in (7). The solution
N −1
in (19) with the − sign results in a small value and the solution in 1
(20) with the + sign results in a large value, which are unrealistic. Â4 M = |r(k)|4 . (24)
N
Therefore, the real estimates with the + sign in (19) and the − k=0
sign in (20) are chosen, i.e. The CRLB for noise variance estimation is given by CRLBn =
  σ 4 /N [8, eq. (33)]. Fig. 4 shows a gap between the two
2 1
Âunr = ÂSMm +  4Â2SMm − 3Â2 M , (21) magnitude-based noise variance estimators and CRLBn . Our
3 3 new estimator σ2 outperforms the conventional σ2
mr when the
  r

2
2 4 2 2 2
SNR is greater than 0 dB.
σr = Â2 M − ÂSMm − ÂSMm  4ÂSMm − 3Â2 M .
3 9 9
(22) V. CONCLUSION
We have addressed the problem of estimating the amplitude
The joint ML amplitude estimator in (21) and the noise variance of a single sinusoid observed in Gaussian noise, based on mea-
estimator in (22) work independently and require only informa- surements of the received signal magnitude alone. While this
tion of the magnitude |r(k)|. estimation problem is simpler when the noise variance is known,
Fig. 3 shows that the MSE performance of the amplitude the joint estimation of the amplitude and the noise variance is
estimator Âunr in (21) is much better than that of the magnitude not difficult also when the noise variance is unknown. Consistent
sample mean estimator ÂSMm without the knowledge of σ 2 at application of the ML estimation principle together with accu-
all SNR values. rate analytical approximations of the likelihood equation leads
In comparison, a conventional noise variance estimator using to simple estimators that show better performance than those of
only knowledge of the received signal magnitude is obtained existing ad hoc estimators and are applicable to fields in [9]–[13]
by approximating the second and fourth absolute moments with where the SNR is typically above 0 dB.
Authorized licensed use limited to: National University Fast. Downloaded on October 31,2024 at 08:48:55 UTC from IEEE Xplore. Restrictions apply.
418 IEEE SIGNAL PROCESSING LETTERS, VOL. 28, 2021

REFERENCES [11] D. Irion Alves et al., “A statistical analysis for wavelength-resolution


SAR image stacks,” IEEE Geosci. Remote Sens. Lett., vol. 17, no. 2,
[1] D. Rife and R. Boorstyn, “Single-tone parameter estimation from discrete- pp. 227–231, Feb. 2020.
time observations,” IEEE Trans. Inf. Theory, vol. IT-20, no. 5, pp. 591–598, [12] W. Wu, H. Guo, and X. Li, “Man-made target detection in urban areas based
Sep. 1974. on a new azimuth stationarity extraction method,” IEEE J. Sel. Topics Appl.
[2] H. Fu and P. Y. Kam, “MAP/ML estimation of the frequency and phase Earth Observ. Remote Sens., vol. 6, no. 3, pp. 1138–1146, Jun. 2013.
of a single sinusoid in noise,” IEEE Trans. Signal Process., vol. 55, no. 3, [13] C. W. Jemmott and R. L. Culver, “Single-hydrophone model-based passive
pp. 834–845, Mar. 2007. sonar source depth classification,” IEEE J. Ocean. Eng., vol. 36, no. 4,
[3] H. Fu and P. Y. Kam, “Phase-based, time-domain estimation of the fre- pp. 696–702, Oct. 2011.
quency and phase of a single sinusoid in AWGN-the role and applications [14] G. Huang, D. Akopian, and C. L. P. Chen, “Measurement and characteriza-
of the additive observation phase noise model,” IEEE Trans. Inf. Theory, tion of channel delays for broadband power line communications,” IEEE
vol. 59, no. 5, pp. 3175–3188, May 2013. Trans. Instrum. Meas., vol. 63, no. 11, pp. 2583–2590, Nov. 2014.
[4] J. G. Proakis, Digital Communications, 4th ed. New York, NY, USA: [15] J. Nielsen and G. F. Pedersen, “Estimation of Rician channels from indoor
McGraw Hill, 2001. measurements at 26 GHz,” in Proc. IEEE 29th Annu. Int. Symp. Pers.,
[5] B. Backus, “Bias due to noise in otoacoustic emissions measurements,” J. Indoor Mobile Radio Commun., Sep. 2018, pp. 1–5.
Acoust. Soc. Ame., vol. 121, no. 3, pp. 1588–603, Mar. 2007. [16] L. Jiang and W. Yang, “Adaptive magnetic resonance image denoising us-
[6] G. McGibney and M. R. Smith, “An unbiased signal-to-noise ratio measure ing mixture model and wavelet,” in Proc. Int. Conf. Digit. Image Comput.:
for magnetic resonance images,” Med. Phys., vol. 20, no. 4, pp. 1077–1078, Techn. Appl., 2003, pp. 831–838.
Jul./Aug. 1993. [17] A. Papoulis and S. U. Pillai, Probability, Random Variables and Stochastic
[7] J. Sijbers, A. J. den Dekker, P. Scheunders, and D. Van Dyck, “Maximum- Processes, 4th ed. New York, NY, USA: McGraw-Hill, 2002.
likelihood estimation of Rician distribution parameters,” IEEE Trans. Med. [18] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products,
Imag., vol. 17, no. 3, pp. 357–61, Jun. 1998. 7th ed. Burlington, MA, USA: Elsevier, 2007.
[8] J. Sijbers and A. J. den Dekker, “Maximum likelihood estimation of signal [19] Y. Jin, T. Meng, S. Zhi, M. Wu, and W. Zhou, “Iterative algorithms for
amplitude and noise variance from MR data,” Magn. Reson. Med., vol. 51, maximum likelihood magnitude-based amplitude estimation,” J. Zhejiang
no. 3, pp. 586–594, Mar. 2004. Univ. Sci. Technol., vol. 32, no. 3, pp. 174–180, Jun. 2020.
[9] M. Bouhrara, J. M. Bonny, B. G. Ashinsky, M. C. Maring, and [20] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions
R. G. Spencer, “Noise estimation and reduction in magnetic resonance With Formulas, Graphs, and Mathematical Tables. New York, NY, USA:
imaging using a new multispectral nonlocal maximum-likelihood filter,” United States Government Printing Office, 1972.
IEEE Trans. Med. Imag., vol. 36, no. 1, pp. 181–193, Jan. 2017.
[10] D. Xie and W. Zhang, “Estimating speech spectral amplitude based on
the Nakagami approximation,” IEEE Signal Process. Lett., vol. 21, no. 11,
pp. 1375–1379, Nov. 2014.

Authorized licensed use limited to: National University Fast. Downloaded on October 31,2024 at 08:48:55 UTC from IEEE Xplore. Restrictions apply.

You might also like