关键词: 核方法, 局部线性嵌入, 支持向量机, 心电图识别

Abstract: This paper investigates kernel method and Locally Linear Embedding（LLE） for proposed kernel method based locally linear embedding features extraction algorithm.The proposed kernel method based locally linear embedding overcomes the instability of locally linear embedding caused by incongruous ECG features distribution.Furthermore，this paper has realized the emulation experiments of MIT-BIH ECG Arrhythmias data base combining with support vector machine.A full comparison of features extraction by PCA and LLE demonstrates that the proposed method is effective.

Key words: kernel method, locally linear embedding, support vector machine, ECG recognition