隐Markov模型的多算法交叉耦合自适应算法

以离子通道信号重构为例，扩展HMM为矢量隐Markov模型，利用随机逼近原理对期望最大算法进行自适应改造，估计离子通道的动力学特征参数；递归辅助变量算法估计背景噪声的统计特征；卡尔曼滤波预测背景噪声；三种算法交叉耦合重构离子通道信号。该算法能够克服滤波器和背景噪声的影响，在低信噪比情况下得到了较高精度的估计参数和重构信号，具有鲁棒性和一致收敛特性。

Adaptive Multi-Algorithm Cross-Coupling Based on Hidden Markov Models

To study the cross-coupling of hidden Markov model (HMM) technology with traditional signal processing algorithms,as the example of the cross-coupling,the ion channel signal is reconstructed and HMM is extended to the vector HMM.Based on the stochastic approximation principle the expectation-maximization(EM) algorithm is improved as the adaptive algorithm.The adaptive EM algorithm used in the vector HMM can estimate the kinetic parameters of the ion channel signal.The recursive instrumental variable algorithm deduces the characteristic of the background noise,and the Kalman filter predictes the background noise.The cross-coupling of the three algorithms reconstructes the ion channel signal.The cross-coupling algorithm estimates the parameters,reconstructes the ion channel signal from the low signal-noise ratio sampled data,and overcomes the side-effect of the patch-clamp system filter and the colored background noise.Besides,it convergences consistently and is robust.