Single trial estimation of event‐related potential (ERP) components is an open research topic in neuroscience. In this article, we have proposed a method to improve the performance of spatiotemporal filtering by decreasing its dependency to prior estimates of ERP components. For this purpose, we have used a mixture of Gaussian kernels instead of a raw prior signal, and the parameters of the Gaussian kernel are estimated using artificial bee colony algorithm. The algorithm starts with one Gaussian kernel, and after optimizing its parameters, another Gaussian kernel is added. This procedure goes on until the stopping criterion is reached. The efficiency of the algorithm is tested for one single uncorrelated component and two correlated components for synthesized electroencephalogram (EEG) signal. Also, the efficiency of the proposed method is presented on real data for extraction of N170 component in real EEG data.

中文翻译：

---

使用时空滤波和人工蜂群优化的高斯核混合物模型对事件相关潜在成分进行单次试验估计

事件相关电位（ERP）组件的单次试验估计是神经科学领域的一个开放研究主题。在本文中，我们提出了一种通过减少时空过滤对ERP组件先前估计的依赖性来提高时空过滤性能的方法。为此，我们使用了混合的高斯核而不是原始先验信号，并且使用人工蜂群算法估计了高斯核的参数。该算法从一个高斯核开始，在优化其参数之后，添加了另一个高斯核。该过程一直进行到达到停止标准为止。针对合成的脑电图（EEG）信号，针对单个不相关分量和两个相关分量测试了算法的效率。也，