基于贝叶斯因子分析模型的激光回波信号增强方法

Laser Echo Waveform Enhancement Method Based on Bayesian Factor Analysis Model

激光回波易被噪声污染,被干扰或遮挡时会出现数据缺失。针对该问题,提出一种基于贝叶斯因子分析(Bayesian factor analysis, BFA)模型的激光回波信号增强算法。首先利用因子分析(factor analysis, FA)模型对激光回波进行建模,从而将信号增强问题转换为对模型参数的求解问题;然后将自动相关确定(automatic relevance determination, ARD)理论引入FA模型,实现对因子个数的自适应寻优;最后采用变分贝叶斯期望最大(variational Bayes expectation maximization, VBEM)算法对模型参数进行迭代更新,在贝叶斯后验概率最大准则下实现激光回波的噪声抑制和缺失样本恢复。基于实测数据开展试验,在低信噪比和数据随机缺失情况下,所提方法能够获得较好的噪声抑制性能和较高的缺失样本恢复性能,从而保证后续信号处理能够准确采集到回波中的高价值信息。

To solve the problem that the laser echo is easily polluted by noise and the sample will be lost randomly when there is occlusion, which will affect the extraction of useful information in the echo, a method of laser echo signal enhancement based on the bayesian factor analysis(BFA) model is proposed. Firstly, based on the analysis of the characteristics of the laser echo, the orthogonal factor model is used to model it, so that the problem of signal enhancement can be transformed into the problem of solving the model parameters. Then the variable bayes expectation maximization(VBEM) algorithm is used to update the model parameters adaptively and iteratively, and the laser echo enhancement can be realized under the minimum reconstruction error criterion. Finally, the measured data are utilized to carry out the experiment. In the environment of low SNR and incomplete random missing data, the proposed method can obtain better noise suppression performance and higher missing sample reconstruction performance, so as to ensure that the follow-up signal processing can accurately collect high-value information in the echo.