Rice distribution explained

In probability theory, the Rice distribution or Rician distribution (or, less commonly, Ricean distribution) is the probability distribution of the magnitude of a circularly-symmetric bivariate normal random variable, possibly with non-zero mean (noncentral). It was named after Stephen O. Rice (1907 - 1986).

Characterization

The probability density function is

f(x\mid\nu,\sigma)=

x\exp\left(
\sigma2
-(x2+\nu2)
2\sigma2
\right)I
0\left(x\nu
\sigma2

\right),

where I0(z) is the modified Bessel function of the first kind with order zero.

In the context of Rician fading, the distribution is often also rewritten using the Shape Parameter

K=

\nu2
2\sigma2
, defined as the ratio of the power contributions by line-of-sight path to the remaining multipaths, and the Scale parameter

\Omega=\nu2+2\sigma2

, defined as the total power received in all paths.[1]

The characteristic function of the Rice distribution is given as:[2] [3]

\begin{align} \chiX(t\mid\nu,\sigma) =\exp\left(-

\nu2
2\sigma2

\right)&\left[\Psi2\left(1;1,

1
2

;

\nu2
2\sigma2

,-

1
2

\sigma2t2\right)\right.\\[8pt] &\left.{}+i\sqrt{2}\sigmat\Psi2\left(

3
2

;1,

3
2

;

\nu2
2\sigma2

,-

1
2

\sigma2t2\right)\right], \end{align}

where

\Psi2\left(\alpha;\gamma,\gamma';x,y\right)

is one of Horn's confluent hypergeometric functions with two variables and convergent for all finite values of

x

and

y

. It is given by:[4] [5]

\Psi2\left(\alpha;\gamma,\gamma';x,y\right)=

infty
\sum
n=0
infty
\sum
m=0
(\alpha)m+n
(\gamma)m(\gamma')n
xmyn
m!n!

,

where

(x)n=x(x+1)(x+n-1)=

\Gamma(x+n)
\Gamma(x)

is the rising factorial.

Properties

Moments

The first few raw moments are:

'
\begin{align} \mu
1

&=\sigma\sqrt{\pi/2}L1/2(-\nu2/2\sigma2)

'
\\ \mu
2

&=2\sigma2+\nu2

'
\\ \mu
3

&=

3\sqrt{\pi/2}L
3\sigma
3/2

(-\nu2/2\sigma2)

'
\\ \mu
4

&=8\sigma4+8\sigma2\nu2+\nu4

'
\\ \mu
5
5\sqrt{\pi/2}L
&=15\sigma
5/2

(-\nu2/2\sigma2)

'
\\ \mu
6

&=48\sigma6+72\sigma4\nu2+18\sigma2\nu4+\nu6 \end{align}

and, in general, the raw moments are given by

'
\mu
k

=\sigmak2k/2\Gamma(1+k/2)Lk/2(-\nu2/2\sigma2).

Here Lq(x) denotes a Laguerre polynomial:

Lq(x)=L

(0)
q

(x)=M(-q,1,x)=1F1(-q;1;x)

where

M(a,b,z)=1F1(a;b;z)

is the confluent hypergeometric function of the first kind. When k is even, the raw moments become simple polynomials in σ and ν, as in the examples above.

For the case q = 1/2:

\begin{align} L1/2(x)&=1F1\left(-

1
2

;1;x\right)\\ &=ex/2

\left[\left(1-x\right)I
0\left(-x
2

\right)

-xI
1\left(-x
2

\right)\right]. \end{align}

The second central moment, the variance, is

\mu2=2\sigma2+\nu2-(\pi\sigma2/2)L

2
1/2

(-\nu2/2\sigma2).

Note that

2
L
1/2

()

indicates the square of the Laguerre polynomial

L1/2()

, not the generalized Laguerre polynomial
(2)
L
1/2

().

Related distributions

R\simRice\left(|\nu|,\sigma\right)

if

R=\sqrt{X2+Y2}

where

X\simN\left(\nu\cos\theta,\sigma2\right)

and

Y\simN\left(\nu\sin\theta,\sigma2\right)

are statistically independent normal random variables and

\theta

is any real number.

R\simRice\left(\nu,\sigma\right)

comes from the following steps:

P

having a Poisson distribution with parameter (also mean, for a Poisson)

λ=

\nu2
2\sigma2

.

X

having a chi-squared distribution with degrees of freedom.

R=\sigma\sqrt{X}.

R\sim\operatorname{Rice}(\nu,1)

then

R2

has a noncentral chi-squared distribution with two degrees of freedom and noncentrality parameter

\nu2

.

R\sim\operatorname{Rice}(\nu,1)

then

R

has a noncentral chi distribution with two degrees of freedom and noncentrality parameter

\nu

.

R\sim\operatorname{Rice}(0,\sigma)

then

R\sim\operatorname{Rayleigh}(\sigma)

, i.e., for the special case of the Rice distribution given by

\nu=0

, the distribution becomes the Rayleigh distribution, for which the variance is

\mu2=

4-\pi
2

\sigma2

.

R\sim\operatorname{Rice}(0,\sigma)

then

R2

has an exponential distribution.[6]

R\sim\operatorname{Rice}\left(\nu,\sigma\right)

then

1/R

has an Inverse Rician distribution.[7]

Limiting cases

For large values of the argument, the Laguerre polynomial becomes[8]

\limx

L
\nu(x)=|x|\nu
\Gamma(1+\nu)

.

It is seen that as ν becomes large or σ becomes small the mean becomes ν and the variance becomes σ2.

The transition to a Gaussian approximation proceeds as follows. From Bessel function theory we have

I\alpha(z)\to

ez
\sqrt{2\piz
} \left(1 - \frac + \cdots \right) \text z \rightarrow \infty so, in the large

x\nu/\sigma2

region, an asymptotic expansion of the Rician distribution:

\begin{align}f(x,\nu,\sigma)={}&

x\exp\left(
\sigma2
-(x2+\nu2)
2\sigma2
\right)I
0\left(x\nu
\sigma2

\right)\\ is\\ &

x\exp\left(
\sigma2
-(x2+\nu2)
2\sigma2

\right)\sqrt{

\sigma2
2\pix\nu
} \exp \left(\frac \right) \left(1 + \frac + \cdots \right)\\\rightarrow & \frac\exp\left(-\frac\right) \sqrt, \;\;\; \text \frac \rightarrow \infty\end

Moreover, when the density is concentrated around \nu and |x - \nu| \ll \sigma because of the Gaussian exponent, we can also write \sqrt \approx 1 and finally get the Normal approximation

f(x,\nu,\sigma)

1
\sigma\sqrt{2\pi
} \exp\left(- \frac\right), \;\;\; \frac \gg 1The approximation becomes usable for
\nu
\sigma

>3

Parameter estimation (the Koay inversion technique)

There are three different methods for estimating the parameters of the Rice distribution, (1) method of moments,[9] [10] [11] [12] (2) method of maximum likelihood,[9] [10] [11] [13] and (3) method of least squares. In the first two methods the interest is in estimating the parameters of the distribution, ν and σ, from a sample of data. This can be done using the method of moments, e.g., the sample mean and the sample standard deviation. The sample mean is an estimate of μ1' and the sample standard deviation is an estimate of μ21/2.

The following is an efficient method, known as the "Koay inversion technique".[14] for solving the estimating equations, based on the sample mean and the sample standard deviation, simultaneously . This inversion technique is also known as the fixed point formula of SNR. Earlier works[9] [15] on the method of moments usually use a root-finding method to solve the problem, which is not efficient.

First, the ratio of the sample mean to the sample standard deviation is defined as r, i.e.,

1/2
r=\mu
2
. The fixed point formula of SNR is expressed as

g(\theta)=\sqrt{\xi{(\theta)}\left[1+r2\right]-2},

where

\theta

is the ratio of the parameters, i.e.,

\theta={\nu}/{\sigma}

, and

\xi{\left(\theta\right)}

is given by:

\xi{\left(\theta\right)}=2+\theta2-

\pi
8

\exp{(-\theta2/2)}\left[(2+\theta2)I0(\theta2/4)+\theta2

2
I
1(\theta

/4)\right]2,

where

I0

and

I1

are modified Bessel functions of the first kind.

Note that

\xi{\left(\theta\right)}

is a scaling factor of

\sigma

and is related to

\mu2

by:

\mu2=\xi{\left(\theta\right)}\sigma2.

To find the fixed point,

\theta*

, of

g

, an initial solution is selected,

{\theta}0

, that is greater than the lower bound, which is

{\theta}lowerbound=0

and occurs when r = \sqrt[14] (Notice that this is the
1/2
r=\mu
2
of a Rayleigh distribution). This provides a starting point for the iteration, which uses functional composition, and this continues until

\left|gi\left(\theta0\right)-\thetai-1\right|

is less than some small positive value. Here,

gi

denotes the composition of the same function,

g

,

i

times. In practice, we associate the final

\thetan

for some integer

n

as the fixed point,

\theta*

, i.e.,

\theta*=g\left(\theta*\right)

.

Once the fixed point is found, the estimates

\nu

and

\sigma

are found through the scaling function,

\xi{\left(\theta\right)}

, as follows:

\sigma=

1/2
\mu
2
\sqrt{\xi\left(\theta*\right)
},

and

\nu=\sqrt{\left(

'~2
\mu
1

+\left(\xi\left(\theta*\right)-2\right)\sigma2\right)}.

To speed up the iteration even more, one can use the Newton's method of root-finding.[14] This particular approach is highly efficient.

Applications

See also

Further reading

External links

Notes and References

  1. Abdi, A. and Tepedelenlioglu, C. and Kaveh, M. and Giannakis, G., "On the estimation of the K parameter for the Rice fading distribution", IEEE Communications Letters, March 2001, p. 92–94
  2. [#refLiu2007|Liu 2007 (in one of Horn's confluent hypergeometric functions with two variables).]
  3. [#refAnnamalai2000|Annamalai 2000 (in a sum of infinite series).]
  4. [#refErdelyi1953|Erdelyi 1953.]
  5. [#refSrivastava1985|Srivastava 1985.]
  6. Richards, M.A., Rice Distribution for RCS, Georgia Institute of Technology (Sep 2006)
  7. Jones, Jessica L., Joyce McLaughlin, and Daniel Renzi. "The noise distribution in a shear wave speed image computed using arrival times at fixed spatial positions.", Inverse Problems 33.5 (2017): 055012.
  8. Abramowitz and Stegun (1968) §13.5.1
  9. [#RefTalukdar|Talukdar et al. 1991]
  10. [#RefBonny|Bonny et al. 1996]
  11. [#RefSijbers|Sijbers et al. 1998]
  12. [#RefDekker2014|den Dekker and Sijbers 2014]
  13. [#RefVaradarajan|Varadarajan and Haldar 2015]
  14. [#refKoay2006|Koay et al. 2006 (known as the SNR fixed point formula).]
  15. [#RefAbdi|Abdi 2001]
  16. Web site: Ballistipedia. 4 May 2014.
  17. Beaulieu. Norman C. Hemachandra. Kasun. September 2011 . Novel Representations for the Bivariate Rician Distribution. IEEE Transactions on Communications. 59. 11. 2951–2954 . 10.1109/TCOMM.2011.092011.090171. 1221747 .
  18. Dharmawansa. Prathapasinghe. Rajatheva. Nandana. Tellambura. Chinthananda. March 2009. New Series Representation for the Trivariate Non-Central Chi-Squared Distribution. IEEE Transactions on Communications. 57 . 3. 665–675. 10.1109/TCOMM.2009.03.070083. 10.1.1.582.533. 15706035 .
  19. Laskar . J. . 2008-07-01 . Chaotic diffusion in the Solar System . Icarus . en . 196 . 1 . 1–15 . 10.1016/j.icarus.2008.02.017 . 0802.3371 . 2008Icar..196....1L . 11586168 . 0019-1035.