基于隐马尔可夫模型平滑估计的随机噪声压制方法

以地震勘探记录去噪为目标，本文提出了一种隐马尔可夫模型平滑估计方法.它是在基本隐马尔可夫模型滤波基础之上，运用信号检测环节将带噪信号段和无信号段加以区分，构建带噪地震记录的状态转移模型，在贝叶斯框架下，利用平滑密度函数进行状态估计，从而达到压制噪声的目的.数值模拟表明，无论对信噪比还是均方误差，隐马尔可夫模型平滑估计处理后的重构信号优于常规的维纳滤波所恢复信号.我们可以期待这种方法会成为实际地震记录噪声压制的有效手段.

Stochastic noise suppression method based on hidden Markov model smoothing estimate

Aiming at denoising seismic prospecting records, this paper proposes a method of Hidden Markov Model smoothing estimate(HMM-S). Based on the basic Hidden Markov Model filter, it plots the whole seismic records into noisy signal segment and no signal segment by signal detection, establishes the state transition model of noisy seismic records, and under Bayesian framework, implements state estimate using smoothing density function to suppress noise. Simulation shows that for SNR and MSE, the status of the recovered signal processed by Hidden Markov Model smoothing estimate(HMM-S) excels those of Wiener filter.