如何形式化定义复随机变量的概率分布？相关问题及实例咨询

Great set of questions—complex random variables are a workhorse in fields like signal processing, probability theory, and digital communications, so getting the formal foundations right is key. Let’s break this down clearly:

A complex random variable ( Z ) is defined as ( Z = X + iY ), where ( X ) and ( Y ) are real-valued random variables living on the same probability space. To formalize its probability distribution, we use the fact that the complex plane ( \mathbb{C} ) is essentially just ( \mathbb{R}^2 ) with a different notation.

Formally, the distribution of ( Z ) is a probability measure ( \mathbb{P}_Z ) on the Borel ( \sigma )-algebra of ( \mathbb{C} ) (the collection of all "nice" subsets of the plane, generated by open rectangles). For any Borel set ( B \subseteq \mathbb{C} ), we define:
[
\mathbb{P}_Z(B) = \mathbb{P}\left( (X,Y) \in \pi^{-1}(B) \right)
]
Here, ( \pi: \mathbb{R}^2 \to \mathbb{C} ) is the straightforward bijection ( \pi(x,y) = x + iy ). In practice, we often use the joint cumulative distribution function (CDF) of ( X ) and ( Y ) to describe ( Z )'s behavior:
[
F_Z(z) = \mathbb{P}\left( X \leq \text{Re}(z), Y \leq \text{Im}(z) \right)
]
for any ( z = x + iy \in \mathbb{C} ). This CDF captures all probabilistic information about ( Z ), since any event involving ( Z ) can be rewritten as an event about the pair ( (X,Y) ).

Absolutely—they’re completely equivalent. The distribution of a complex random variable ( Z = X + iY ) is in one-to-one correspondence with the joint distribution of the real pair ( (X,Y) ).

Here’s the breakdown:

• If you have ( Z ), you can extract its real and imaginary parts ( X = \text{Re}(Z) ) and ( Y = \text{Im}(Z) ), and their joint distribution uniquely determines every probability involving ( Z ) (as shown in the first question).
• Conversely, any joint distribution for a bivariate real random variable ( (X,Y) ) defines a complex random variable ( Z = X + iY ), whose distribution is exactly the one induced by mapping ( (X,Y) ) to the complex plane.

This equivalence is why we often treat complex random variables as "pairs of real random variables" when doing formal probability, but the complex notation makes it much cleaner to work with in applications like signal processing.

Moments of complex random variables use a complex-valued expectation operator, which preserves linearity even for complex scalars (meaning ( \mathbb{E}[aZ + bW] = a\mathbb{E}[Z] + b\mathbb{E}[W] ) for any complex constants ( a,b )).

For ( Z = X + iY ):

• The first moment (mean) is simple: ( \mathbb{E}[Z] = \mathbb{E}[X] + i\mathbb{E}[Y] ). This is just combining the means of the real and imaginary parts.
• Higher-order moments are defined as ( \mathbb{E}[Z^k] ) for integer ( k \geq 1 ). Using the binomial theorem, we expand ( Z^k = (X + iY)^k ) into terms involving ( X^m Y^{k-m} ) multiplied by complex coefficients. For example, the second moment:
  [
  \mathbb{E}[Z^2] = \mathbb{E}[(X + iY)^2] = \mathbb{E}[X^2 - Y^2] + i\mathbb{E}[2XY] = \left( \mathbb{E}[X^2] - \mathbb{E}[Y^2] \right) + 2i\mathbb{E}[XY]
  ]
  Here, we split the complex expectation into real and imaginary parts, each of which relies on moments or cross-moments of ( X ) and ( Y ).
• We also frequently use mixed moments involving the conjugate of ( Z ), like ( \mathbb{E}[Z^k \overline{Z}^m] ) (critical for autocorrelation in signal processing). For example, the average power of ( Z ) is ( \mathbb{E}[|Z|^2] = \mathbb{E}[Z \overline{Z}] = \mathbb{E}[X^2 + Y^2] = \mathbb{E}[X^2] + \mathbb{E}[Y^2] ).

So to answer your question: Yes, you compute complex moments by expanding the complex expression, calculating the necessary real/imaginary part moments and cross-moments, then combining them with the appropriate complex coefficients. It’s not just arbitrary splitting—it’s leveraging the linearity of expectation to decompose complex moments into real-valued calculations we already know how to do.

One of the most common real-world uses is QPSK (Quadrature Phase Shift Keying), a modulation scheme used in Wi-Fi, 4G/5G, and satellite communications. Here’s how the formal definitions come to life:

• Signal Model: Each transmitted symbol is a complex random variable ( Z = X + iY ). ( X ) and ( Y ) are independent real random variables, each taking values ( +1/\sqrt{2} ) or ( -1/\sqrt{2} ) with equal probability (1/2). This means ( Z ) can be one of four points in the complex plane: ( \pm 1/\sqrt{2} \pm i/\sqrt{2} ), each with probability 1/4.
• Key Moment Calculations:
  • Mean: ( \mathbb{E}[Z] = \mathbb{E}[X] + i\mathbb{E}[Y] = 0 + i0 = 0 ) (zero DC offset, which is standard to avoid wasting power on constant signals).
  • Average Power: ( \mathbb{E}[|Z|^2] = \mathbb{E}[X^2 + Y^2] = (1/2) + (1/2) = 1 ) (normalized power, which simplifies system design and comparisons).
• Noise Model: The channel adds complex Gaussian noise ( N = N_I + iN_Q ), where ( N_I ) and ( N_Q ) are independent real Gaussian variables with mean 0 and variance ( \sigma^2/2 ). This is called a circularly symmetric complex Gaussian (CSCG) variable, denoted ( N \sim \mathcal{CN}(0, \sigma^2) ).
• Error Rate Calculation: To find the probability that the receiver misinterprets a symbol, we analyze the distribution of ( Z + N ). We calculate the probability that ( Z + N ) falls into the wrong decision region in the complex plane—this directly uses the joint distribution of ( N_I ) and ( N_Q ) (i.e., the complex distribution of ( N )) and the moments we computed for ( Z ).

This example shows how the formal definitions of complex random variables and their moments aren’t just abstract math—they’re the foundation for designing and analyzing real-world communication systems.

内容的提问来源于stack exchange，提问作者user188529