Scribd
Upload
207 views6 pages

Mean and Autocorrelation of Filtered Process

1) A random process passed through a linear filter produces a random output process. 2) The mean and autocorrelation of the output process can be estimated based on the input process and filter characteristics. 3) If the input is wide-sense stationary, the power spectral density of the input determines the mean square value of the output.

Uploaded by

Harsha
Copyright
© Attribution Non-Commercial (BY-NC)
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
207 views6 pages

Mean and Autocorrelation of Filtered Process

1) A random process passed through a linear filter produces a random output process. 2) The mean and autocorrelation of the output process can be estimated based on the input process and filter characteristics. 3) If the input is wide-sense stationary, the power spectral density of the input determines the mean square value of the output.

Uploaded by

Harsha
Copyright
© Attribution Non-Commercial (BY-NC)
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
  • Random Process through a Linear Filter

' $

Random Process through a linear filter


• A random process X(t) is applied as input to a linear
time-invariant filter of impulse response h(t),
• It produces a random process Y (t) at the filter output as
shown in Figure 1

X(t) Y(t)
h(t)

Figure 1: Transmission of a random process through a linear filter

• Difficult to describe the probability distribution of the output


random process Y (t), even when the probability distribution of
the input random process X(t) is completely specified for

& %
−∞ ≤ t ≤ +∞.
' $
• Estimate characteristics like mean and autocorrelation of the
output and try to analyse its behaviour.
• Mean The input to the above system X(t) is assumed
stationary. The mean of the output random process Y (t) can
be calculated

»Z +∞ –
mY (t) = E[Y (t)] = E h(τ )X(t − τ ) dτ
−∞
Z +∞
= h(τ )E[X(t − τ )] dτ
−∞
Z +∞
= h(τ )mX (t − τ ) dτ
−∞
Z +∞
= mX h(τ ) dτ
−∞
= mX H(0)

& %
where H(0) is the zero frequency response of the system.
' $
• Autocorrelation The autocorrelation function of the output
random process Y (t). By definition, we have

RY (t, u) = E[Y (t)Y (u)]

where t and u denote the time instants at which the process is


observed. We may therefore use the convolution integral to
write

»Z +∞ Z +∞ –
RY (t, u) = E h(τ1 )X(t − τ1 ) dτ1 h(τ2 )X(t − τ2 ) dτ2
−∞ −∞
Z +∞ Z +∞
= h(τ1 ) dτ1 h(τ2 )E [X(t − τ1 )X(t − τ2 )] dτ2
−∞ −∞

• When the input X(t) is a wide-stationary random process,


– The autocorrelation function of X(t) is only a function of

& %
the difference between the observation times t − τ1 and
' $

u − τ2 .
– Putting τ = t − u, we get

Z +∞ Z +∞
RY (τ ) = h(τ1 )h(τ2 )RX (τ − τ1 + τ2 ) dτ1 dτ2
−∞ −∞

– RY (0) = E[Y 2 (t)]


– The mean square value of the output random process Y (t)
is obtained by putting τ = 0 in the above equation.
Z +∞ Z +∞
2
E[Y (t)] = h(τ1 )h(τ2 )RX (τ2 − τ1 ) dτ1 dτ2
−∞ −∞
" #
1 Z +∞ Z +∞ Z +∞
= H(ω) exp(jωτ1 ) dω h(τ2 )RX (τ2 − τ1 ) dτ1 dτ2
2π −∞ −∞ −∞
1 Z +∞ Z +∞ Z +∞
= H(ω) dω h(τ2 ) dτ2 RX (τ2 − τ1 ) exp(j2ωτ1 ) dτ1
2π −∞ −∞ −∞

– Putting τ = τ2 − τ1

& %
' $

2 1 Z +∞ Z +∞ Z +∞
E[Y (t)] = H(ω) dω h(τ2 ) exp(jωτ2 ) dτ2 RX (τ ) exp(−j2ωτ ) dτ
2π −∞ −∞ −∞
1 Z +∞ Z +∞ Z +∞
= H(ω) dω H ∗ (ω) dω RX (τ ) exp(−jωτ ) dτ
2π −∞ −∞ −∞

– This is simply the Fourier Transform of the autocorrelation


function RX (t) of the input random process X(t). Let this
transform be denoted by SX (f ).

Z +∞
SX (ω) = RX (τ ) exp(−jωτ ) dτ
−∞

– SX (ω) is called the power spectral density or power spectrum


of the wide-sense stationary random process X(t).

Z +∞
2 1
E[Y (t)] = |H(ω)|2 SX (ω) df
2π −∞

& %
' $

– The mean square value of the output of a stable linear


time-invariant filter in response to a wide-sense stationary
random process is equal to the integral over all frequencies
of the power spectral density of the input random process
multiplied by the squared magnitude of the transfer function
of the filter.

& %

Common questions

Powered by AI

The Fourier Transform is used to express the autocorrelation function RX(τ) as the power spectral density SX(ω). This is done through the integral SX(ω) = ∫ RX(τ) exp(-jωτ) dτ, where the transform provides a frequency domain representation of the autocorrelation function, highlighting how different frequency components contribute to the power of the random process .

The mean of the output random process Y(t) can be calculated using the formula mY(t) = mX H(0), where mX is the mean of the input process X(t), and H(0) is the zero frequency response of the system. This implies that the mean of the output process depends linearly on the mean of the input process and the filter's zero frequency response .

The zero frequency response H(0) plays a critical role in determining the mean of the output from a stationary random process passed through a linear filter. It represents the filter's gain at zero frequency (DC component), which directly scales the mean of the input process to produce the mean of the output, demonstrating the influence of the filter's baseline gain characteristics in stationary processes .

For a wide-sense stationary random process, the autocorrelation function of the output process Y(t), RY(τ), is formulated by convolving the filter's impulse response with the autocorrelation function of the input process X(t). Specifically, it is given by RY(τ) = ∫∫ h(τ1) h(τ2) RX(τ - τ1 + τ2) dτ1 dτ2, indicating that it's a function of the relative time difference τ, reflecting the stationary nature of the process .

The filter's impulse response function h(t) affects the autocorrelation RY(t, u) of the output by dictating how the input's autocorrelation is convolved over time. This convolution influences how correlation between different time points in the input translates into the output, determined by the integral RY(τ) = ∫∫ h(τ1) h(τ2) RX(τ - τ1 + τ2) dτ1 dτ2, where the filter's time-invariant properties define the bounds and nature of this convolution .

To derive the mean of the output process Y(t), the expectation operation is applied to the convolution integral of the filter's impulse response and the input process: mY(t) = E[∫ h(τ)X(t − τ) dτ] simplifies to mY(t) = mX ∫ h(τ) dτ, using the stationarity of X(t) and the linearity of expectation. This involves integration over the entire time domain and exploits the input's stationarity .

The squared magnitude of the transfer function |H(ω)|^2 is significant because it represents the frequency response of the filter, affecting how energy at each frequency is attenuated or amplified. In the context of random processes, this squared magnitude determines how the power spectral density of the input influences the mean square value of the output, effectively shaping the spectral content of the output process .

Describing the probability distribution of the output process Y(t) is challenging even if the input distribution is specified because the filter's action can significantly alter the statistical dependencies and characteristics, especially under non-linear or complex filters. Additionally, the convolution operations and integrals involved in filter processes often result in complex transformations of the input characteristics, complicating direct probabilistic descriptions .

The mean square value of the output of a stable linear time-invariant filter is determined by the integral over all frequencies of the power spectral density of the input random process multiplied by the squared magnitude of the filter's transfer function. This relationship shows the influence of both the input spectral characteristics and the filter's frequency response on the output mean square value .

The stability of a linear time-invariant filter ensures that the mean square value of the output process is finite and can be calculated as the integral over all frequencies of the power spectral density of the input process multiplied by the squared magnitude of the filter's transfer function. Stability implies bounded outputs for bounded inputs, crucial for meaningful calculation of output statistics .

' & $ % Random Process through a linear filter • A random process X(t) is applied as input to a linear time-invariant filter of
' & $ % • Estimate characteristics like mean and autocorrelation of the output and try to analyse its behaviour. • Mean The i
' & $ % • Autocorrelation The autocorrelation function of the output random process Y (t). By definition, we have RY (t, u) =
' & $ % u −τ2. – Putting τ = t −u, we get RY (τ) = Z +∞ −∞ Z +∞ −∞ h(τ1)h(τ2)RX(τ −τ1 + τ2) dτ1 dτ2 – RY (0) = E[Y 2(t)] – Th
' & $ % E[Y 2(t)] = 1 2π Z +∞ −∞ H(ω) dω Z +∞ −∞ h(τ2) exp(jωτ2) dτ2 Z +∞ −∞ RX (τ) exp(−j2ωτ) dτ = 1 2π Z +∞ −∞ H(ω) dω Z +∞
' & $ % – The mean square value of the output of a stable linear time-invariant filter in response to a wide-sense stationary

You might also like