Stochastic Process_3平稳过程.pptVIP

随机过程 Stochastic Process 平稳过程 （一类统计特效不随时间推移而变化的随机过程） 基本概念 数字特征 性质 基本概念 严平稳过程：设一随机过程 ，对任意实数 ，n维随机变量 和 有相同联合分布函数(joint distribution)，即 则称该随机过程为严(强)平稳过程，或称其具有严平稳性。 严平稳过程的有限维分布不随时间的推移而变化。 一维分布函数： 与t无关 二维分布函数：仅与时间间隔有关，与两个时刻本身无关 二阶矩严平稳过程 均值函数 相关函数 (宽)平稳过程：设 为二阶矩过程，若 (1) 对任意t， ； (2)对任意s, t， 或 则称该随机过程为(宽/弱)平稳过程. 严平稳过程 vs.宽平稳过程 严平稳过程 宽平稳过程 严平稳过程: 只涉及有限维分布，不要求一、二阶矩的存在 宽平稳过程：只要均值函数与时间无关，相关函数仅与时间间隔有关，与时间起始点无关。推导不出有限维分布不随时间推移而改变。 宽平稳过程不一定是严平稳过程。 定理1： 若 是正态过程，则其为严平稳过程的充要条件是 为宽平稳过程。 证明：充分性：正态 ?二阶矩 必要性： Exp1: We have Let us suppose that the amplitude A is random with no assumption about the PDF of A. we also suppose that the phase is uniformly in the interval . And, A and are independent. Prove the random sinusoid is a WSS process. Verification (1): For each fixed t, we have Verification (2): For each fixed t1 t2, we have which depends only on t1-t2 A Filtering Example Let Zn be the discrete time white noise process with unit variance. This process satisfies Let us pass Zn through a FIR linear filter which does the following The output random signal Xn turns out to be WSS. Tap weights First, Concern with condition (2) The autocorrelations RX(n1,n2) clearly only depend on n1-n2. Properties of 1. For any autocorrelation function , the value RX(0) is nonnegative and The peak value of an autocorrelation function is always RX(0). 2. Any autocorrelation function is an even function of 3. Any autocorrelation function is called the positive semidefiniteness property. Positive Semidefiniteness Property Properties of 4. If the realizations of a WSS process X are all periodic with the same period T, then is also periodic with period T. 5. Construct autocor