连续信源与无记忆Gaussian信源的比较及若干强偏差定理

The Comparison Between Continuous Information Source and Memoryless Gaussian Information Source and Some Strong Deviation Theorems

引入相对熵密度偏差作为一般连续信源相对无记忆Ｇａｕｓｓｉａｎ信源的偏差的一种随机性度量，利用似然比构造几乎处处收敛的上鞅，结合文献［１，４］中提出的分析方法，得到了任意连续型随机变量序列平方和的一类强偏差定理，其中包含连续信源相对熵密度的若干极限定理。

The notion of deviation of relatuve entropy density as the random measure of deviation for general continuous entropy relative to memoryless Gaussian entropy is introduced, we obtain a class of strong deviation theorems of the sum of squar for arbitrary continuous random variabled by using the concept supper martingale and te analytical method proposed in [1] and [4], and some limit theorems of relative entropy density of continuous entropy is included.