Processing algorithm of Brain waves?

一、算法分类（1）信号预处理算法：独立成分分析（ICA）、伪迹减法、小波变换、空间滤波；（2）特征提取算法：时域分析法、频域分析法、时频分析法、空间分析法；（3）信号分类算法：线性判别分析（LDA）、支持向量机（SVM）、k-最邻近法则（KNN）、深度神经网络（DNN）、人工神经网络（ANN）、卷积神经网络（CNN）、极限学习机（GELM）、组合深度学习神经网络（RBM-GRNN）；二、算法详解 1、Continuous homology algorithm is a common tool in topological data analysis. It is an algebraic approach that describes the characteristics of relationships between datasets by visualizing datasets as topological diagrams. These features can help us understand the organizational structure of data sets for better data analysis and mining.(持续同源性算法是拓扑数据分析中的一种常用工具。它是一种代数方法，通过将数据集可视化为拓扑图，用于描述数据集之间关系的特征。这些特征可以帮助我们理解数据集之间的组织结构，从而更好地进行数据分析和挖掘。) 2、Signal processing technology restricts the development of BCI system. The key of signal processing technology lies in feature extraction and classification recognition algorithm. The feature extraction algorithms that have been studied and applied mainly include time domain analysis. (Amplitude Analysis, Waveform Analysis, Coherent Average Analysis) Other analysis methods such as frequency domain analysis (Fourier transform, power spectrum), time-frequency analysis (continuous wavelet transform, empirical mode decomposition) and common space mode (CSP). At present, the main classification and recognition methods are linear discriminant analysis (LDA), regularized LDA, support vector machine (SVM) and naive Bayes classifier. Although these algorithms are proved to be effective under certain conditions, the problems of BCI system, such as large individual difference, small training set and low signal-to-noise ratio, have not been solved fundamentally. These problems affect the classification accuracy, information transfer rate (ITR) and other important technical indexes of BCI system. Therefore, scholars need to continuously study and improve the robust and more complex electroencephalogram signal decoding algorithm to realize the high-level and accurate processing of EEG signals.(信号处理技术制约BCI系统的发展。信号处理技术的关键在于特征提取和分类识别算法。已经研究应用的特征提取算法主要包括时域分析（幅值分析、波形分析、相干平均分析）、频域分析（傅里叶变换、功率谱）、时频分析（连续小波变换、经验模态分解）及共空间模式（CSP）等其他分析方法。目前应用的主要分类识别方法主要有线性判别分析（LDA）、正则化 LDA、支持向量机（SVM）和朴素贝叶斯分类器等。尽管这些算法被证明在一定条件下是行之有效的，但BCI系统存在的被试者个体差异大、训练集小、信噪比低等问题仍未得到根本性解决，这些问题影响着BCI系统的分类准确率、信息传输率（ITR）等重要技术指标。这需要学者们不断研究和改进鲁棒性和复杂更高的脑电信号解码算法，以实现对脑电信号的高校、准确处理。)

Independent component analysis (ICA) method-data preprocessing, using some statistical information to separate the observed signal into a linear combination of multiple independent non-Gaussian source signals. In recent years, independent component analysis has been widely used in many fields, such as wireless communication, image enhancement, biomedical signal processing and so on.(独立成分分析（ICA）方法-数据预处理，利用一些统计信息将观测信号分离为多个独立的非高斯源信号的线性组合。近些年来，独立成分分析方法在诸多领域都有着广泛的应用，例如无线通讯、图像增强、生物医学信号处理等。)
feature extraction: Common Spatial Pattern (CSP) algorithm: The spatial filtering algorithm can extract spatial information. After data is projected, the variance of one type of data is maximized and the variance of another type of data is minimized. Finally, the features representing the two types of data are extracted and combined as the final features. The mechanism of CSP algorithm is based on the event-related de-synchronization/event-related synchronization phenomenon in the brain's motion imagination process, so the effectiveness of CSP algorithm is proved from the side.(特征提取：公共空间模式（CSP）算法-空间滤波算法能够提取空间信息，通过寻找一个隐藏投影空间使得数据投影之后极大化一类数据的方差同时极小化另一类数据的方差，最终分别提取表征两类数据的特征并拼接一起作为最终特征。CSP 算法的机理是基于大脑运动想象过程的事件相关去同步/事件相关同步现象，因此从侧面证明了 CSP 算法的有效性。)
Support Vector Machine (SVM)(支持向量机)。

cmy2022 / neural-interface

Processing algorithm of Brain waves? #13