Cyclostationary process
≈ 8 min read
Names in other languages
No names in other languages are held for this entry yet.
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
A cyclostationary process is a signal having statistical properties that vary cyclically with time. A cyclostationary process can be viewed as multiple interleaved stationary processes. For example, the maximum daily temperature in New York City can be modeled as a cyclostationary process: the maximum temperature on July 21 is statistically different from the temperature on December 20; however, it is a reasonable approximation that the temperature on December 20 of different years has identical statistics. Thus, we can view the random process composed of daily maximum temperatures as 365 interleaved stationary processes, each of which takes on a new value once per year.
A cyclostationary process is a signal having statistical properties that vary cyclically with time. A cyclostationary process can be viewed as multiple interleaved stationary processes. For example, the maximum daily temperature in New York City can be modeled as a cyclostationary process: the maximum temperature on July 21 is statistically different from the temperature on December 20; however, it is a reasonable approximation that the temperature on December 20 of different years has identical statistics.
Thus, we can view the random process composed of daily maximum temperatures as 365 interleaved stationary processes, each of which takes on a new value once per year. R_x(t,\tau) = \operatorname{E} \{ x(t + \tau) x^*(t) \},\, where the star denotes complex conjugation, is said to be wide-sense cyclostationary with period T_0 if both \operatorname{E}[x(t)] and R_x(t,\tau) are cyclic in t with period T_0, i.e.: \operatorname{E}[x(t)] = \operatorname{E}[x(t+T_0)]\text{ for all }t R_x(t,\tau) = R_x(t+T_0; \tau)\text{ for all }t, \tau.
The autocorrelation function is thus periodic in t and can be expanded in Fourier series: R_x(t,\tau) = \sum_{n=-\infty}^\infty R_x^{n/T_0}(\tau) e^{j2\pi\frac{n}{T_0}t} where R_x^{n/T_0}(\tau) is called cyclic autocorrelation function and equal to: R_x^{n/T_0}(\tau) = \frac{1}{T_0} \int_{-T_0/2}^{T_0/2} R_x(t,\tau)e^{-j2\pi\frac{n}{T_0}t} \mathrm{d}t . The frequencies n/T_0,\,n\in \mathbb{Z}, are called cycle frequencies. Wide-sense stationary processes are a special case of cyclostationary processes with only R_x^0(\tau)\ne 0 .
S_x(f, \alpha) = \lim_{T \to \infty} \frac{1}{T} \mathbb{E} \left[ X_T \left( f + \frac{\alpha}{2} \right) X_T^* \left( f - \frac{\alpha}{2} \right) \right] . \widehat{R}_x^{n/T_0}(\tau) = \lim_{T\rightarrow +\infty} \frac{1}{T} \int_{-T/2}^{T/2} x(t + \tau) x^*(t) e^{-j2\pi\frac{n}{T_0}t} \mathrm{d}t . If the time-series is a sample path of a stochastic process it is R_x^{n/T_0}(\tau) =\operatorname{E}\left[\widehat{R}_x^{n/T_0}(\tau)\right] .
If the signal is further cycloergodic, all sample paths exhibit the same cyclic time-averages with probability equal to 1 and thus R_x^{n/T_0}(\tau) =\widehat{R}_x^{n/T_0}(\tau) with probability 1. S_x^\alpha(f) = \int_{-\infty}^{+\infty} R_x^{\alpha}(\tau) e^{-j2\pi f\tau}\mathrm{d}\tau . The cyclic spectrum at zero cyclic frequency is also called average power spectral density. For a Gaussian cyclostationary process, its rate distortion function can be expressed in terms of its cyclic spectrum.
The reason S_x^\alpha(f) is called the spectral correlation density function is that it equals the limit, as filter bandwidth approaches zero, of the expected value of the product of the output of a one-sided bandpass filter with center frequency f + \alpha /2 and the conjugate of the output of another one-sided bandpass filter with center frequency f - \alpha /2 , with both filter outputs frequency shifted to a common center frequency, such as zero, as originally observed and proved in.
For time series, the reason the cyclic spectral density function is called the spectral correlation density function is that it equals the limit, as filter bandwidth approaches zero, of the average over all time of the product of the output of a one-sided bandpass filter with center frequency f + \alpha /2 and the conjugate of the output of another one-sided bandpass filter with center frequency f - \alpha /2 , with both filter outputs frequency shifted to a common center frequency, such as zero, as originally observed and proved in. x(t) = \sum_{k=-\infty}^{\infty} a_k p(t -kT_0) where a_k\in\mathbb{C} are i.i.d. random variables.
The waveform p(t) , with Fourier transform P(f) , is the supporting pulse of the modulation. By assuming \operatorname{E}[a_k] = 0 and \operatorname{E}[|a_k|^2]=\sigma_a^2 , the auto-correlation function is: \begin{align} R_x(t,\tau) &= \operatorname{E}[x(t+\tau)x^*(t)] \\[6pt] &= \sum_{k,n}\operatorname{E}[a_k a_n^*]p(t+\tau-kT_0)p^*(t-nT_0) \\[6pt] &= \sigma_a^2\sum_{k}p(t+\tau-kT_0)p^*(t-kT_0) . \end{align} The last summation is a periodic summation, hence a signal periodic in t.
This way, x(t) is a cyclostationary signal with period T_0 and cyclic autocorrelation function: \begin{align} R_x^{n/T_0}(\tau) &= \frac{1}{T_0}\int_{-T_0/2}^{T_0/2} R_x(t,\tau) e^{-j2\pi\frac{n}{T_0}t} \, \mathrm{d}t \\[6pt] &= \frac{1}{T_0}\int_{-T_0/2}^{T_0/2} \sigma_a^2\sum_{k=-\infty}^\infty p(t+\tau-kT_0)p^*(t-kT_0) e^{-j2\pi\frac{n}{T_0}t}\mathrm{d}t \\[6pt] &= \frac{\sigma_a^2}{T_0} \sum_{k=-\infty}^\infty\int_{-T_0/2-kT_0}^{T_0/2-kT_0}p(\lambda+\tau)p^*(\lambda) e^{-j2\pi\frac{n}{T_0}(\lambda+kT_0)}\mathrm{d}\lambda \\[6pt] &= \frac{\sigma_a^2}{T_0} \int_{-\infty}^\infty p(\lambda+\tau)p^*(\lambda) e^{-j2\pi\frac{n}{T_0}\lambda}\mathrm{d}\lambda \\[6pt] &= \frac{\sigma_a^2}{T_0} p(\tau) * \left\{p^*(-\tau)e^{j2\pi\frac{n}{T_0}\tau}\right\} . \end{align} with * indicating convolution.
The cyclic spectrum is: S_x^{n/T_0}(f) = \frac{\sigma_a^2}{T_0} P(f)P^*\left(f-\frac{n}{T_0}\right) . Typical raised-cosine pulses adopted in digital communications have thus only n=-1, 0, 1 non-zero cyclic frequencies. This same result can be obtained for the non-stochastic time series model of linearly modulated digital signals in which expectation is replaced with infinite time average, but this requires a somewhat modified mathematical method as originally observed and proved in.
Cyclostationary models
It is possible to generalise the class of autoregressive moving average models to incorporate cyclostationary behaviour. For example, Troutman treated autoregressions in which the autoregression coefficients and residual variance are no longer constant but vary cyclically with time. His work follows a number of other studies of cyclostationary processes within the field of time series analysis.
Polycyclostationarity
In practice, signals exhibiting cyclicity with more than one incommensurate period arise and require a generalization of the theory of cyclostationarity. Such signals are called polycyclostationary if they exhibit a finite number of incommensurate periods and almost cyclostationary if they exhibit a countably infinite number. Such signals arise frequently in radio communications due to multiple transmissions with differing sine-wave carrier frequencies and digital symbol rates. The theory was introduced in for stochastic processes and further developed in for non-stochastic time series.
Higher Order and Strict Sense Cyclostationarity
The wide sense theory of time series exhibiting cyclostationarity, polycyclostationarity and almost cyclostationarity originated and developed by Gardner was also generalized by Gardner to a theory of higher-order temporal and spectral moments and cumulants and a strict sense theory of cumulative probability distributions. The encyclopedic book comprehensively teaches all of this and provides a scholarly treatment of the originating publications by Gardner and contributions thereafter by others.
R_x(\theta,\tau) = \operatorname{E} \{ x(t(\theta) + \tau) x^*(t(\theta)) \},\, where \theta stands for angle, t(\theta) for the time instant corresponding to angle \theta and \tau for time delay. Processes whose angle-time autocorrelation function exhibit a component periodic in angle, i.e. such that R_x(\theta;\tau) has a non-zero Fourier-Bohr coefficient for some angular period \Theta , are called (wide-sense) angle-time cyclostationary.
The double Fourier transform of the angle-time autocorrelation function defines the order-frequency spectral correlation, S_x^\alpha(f) = \lim_{S\rightarrow +\infty} \frac{1}{S} \int_{-S/2}^{S/2}\int_{-\infty}^{+\infty} R_x(\theta,\tau) e^{-j2\pi f\tau} e^{-j2\pi\alpha\frac{\theta}{\Theta}} \, \mathrm{d}\tau \, \mathrm{d}\theta where \alpha is an order (unit in events per revolution) and f a frequency (unit in Hz). For constant speed of rotation, \omega , angle is proportional to time, \theta = \omega t .
Consequently, the angle-time autocorrelation is simply a cyclicity-scaled traditional autocorrelation; that is, the cycle frequencies are scaled by \omega . On the other hand, if the speed of rotation changes with time, then the signal is no longer cyclostationary (unless the speed varies periodically). Therefore, it is not a model for cyclostationary signals. It is not even a model for time-warped cyclostationarity, although it can be a useful approximation for sufficiently slow changes in speed of rotation.
Text derived from https://en.wikipedia.org/w/index.php?oldid=1319210981 under CC BY-SA 4.0; modifications have been made.