共模误差PCA与ICA提取方法的比较

为消除共模误差CME，目前广泛应用的主成分分析法（PCA）基于观测数据的二阶统计量（方差和协方差）将观测网残差分解成一组不相关的随时间变化的模态和对应的空间响应，而GPS时间序列分布具有非高斯特征，二阶统计量不能完全描述其随机特性。本文假设区域网CME 与其他误差相互统计独立，则可以采用独立分量分析（ICA）法。采用模拟数据对ICA提取CME的精确性和有效性进行验证，并与PCA结果进行对比。结果表明，ICA能够有效地提取观测网CME。

Analysis and Comparison of Common Mode Error Extraction Using Principal Component Analysis and Independent Component Analysis

Common mode error(CME), a major source of error correlated spatially in regional GPS solutions, should be removed to enhance signal-to-noise ratio in GPS coordinate time series. Principal component analysis (PCA), which is widely used for CME extraction, decomposes the time series of the GPS network into a group of modes, where each mode consists of a common temporal function and corresponding spatial response based on second-order statistics. Since the probability distribution function of GPS time series is sometimes no-Gaussian, the second-order statistic cannot fully capture its stochastic characteristics. In this paper, we assume that CME is stochastic independent with other error sources, so an independent component analysis (ICA) is introduced to analyze it. The performance of ICA is validated and compared with that of PCA through a simulated example.