A principal consideration on general Bayes'filters for a state estimation of environmental sound and vibration systems

This paper describes a fundamental consideration on our works on the design of general Bayes' filters for the state estimation of non-stationary, non-linear, and non-Gaussian environmental sound and vibration systems. We have discussed an essential point of several Bayes' filters proposed by using the orthogonal or non-orthogonal expansion form of Bayes' theorem. They can estimate any kinds of statistics of arbitrary function type of state variables including the lower and the higher order statistics connected with the Lx evaluation index in the environmental sound and vibration systems. Here, we have mainly focussed on giving the fundamental viewpoints of their design policies. Some new estimation methods and new results not yet published are included.

未知量测噪声分布下的多扩展目标CBMeMBer滤波算法

在实际应用场景中,量测噪声协方差准确模型很难被建立,传统的多扩展目标跟踪算法在量测噪声协方差未知情况下跟踪性能迅速下降。为了解决量测噪声未知对多扩展目标跟踪结果造成的影响,将变分贝叶斯方法引入到CBMeMBer滤波算法中。VB-GM-CBMeMBer算法能在量测噪声未知情况下通过估计噪声协方差进行滤波计算,但该算法存在目标数目估计不准确的问题。针对此问题,提出一种改进的VB-GM-CBMeMBer算法,该算法在滤波算法预测步骤后引入椭球门限,使用保留在门限内的量测来进行下一步计算,以减少杂波量测,降低杂波量测对扩展目标量测的影响,提高对扩展目标状态聚类的精度。实验结果表明,该算法适用于多扩展目标数目未知、量测噪声协方差未知的情况,且其跟踪精度比GM-CBMeMBer和VB-GM-CBMeMBer滤波算法有一定提高。