【摘要】 基于稀疏k-means非监督学习的聚类算法,就心律失常、充血性心力衰竭、心肌缺血、突发性心脏死亡及健康心脏电信号进行了分类研究。相比传统k-means算法,非监督学习的聚类算法能将数据从RAM中直接加载并分类,有效节省了分类时间和内存。通过优化稀疏k-means算法中分类输出的迭代方法,构建了有望应用于人体的分类器心脏检测仪。实验表明,经优化的稀疏k-means算法在截取时间为6 s时,处理数据时间短至0.34 s,精确度高达98.52%。并利用Silhouette侧影聚类,对优化后的稀疏k-means算法进行分类校验,验证了算法的有效性,为心脏健康状况实时快速精确监测提供了新思路。更多还原

【Abstract】 The heart conditions such as electrical signals of arrhythmia,congestive heart failure,myocardial ischemia and sudden cardiac death were investigated by using healthy condition as control in terms of category based on sparse k-means unsupervised learning clustering algorithm in the paper.Different to the traditional k-means algorithm,the sparsified k-means algorithm based on unsupervised learning acquired and classified data directly in RAM and sped up data processing and saved classifying time and memory.The classification output of sparse k-means algorithm was optimized by the iterative method and the optimized algorithm was applied to make a heart detector and monitor potential to be used for classifying the heart conditions.The results found that the classification time was 0.34 s and the highest accuracy was 98.52% when the interception time was 6 s with the algorithm optimized.The effectiveness of optimized k-means algorithm confirmed by Silhouette clustering demonstrated that our design could provide a promising route for making a real-time,rapid and accurate heart monitor.更多还原