Arrhythmia Classification Algorithm based on Sparse Autoencoder

Cardiovascular diseases are the leading cause of death from noncommunicable diseases worldwide, among which arrhythmias are a common manifestation. Feature extraction is an important part of arrhythmia classification algorithms. Most traditional classification algorithms rely on manual design and extraction of features. In order to improve the efficiency of feature extraction and reduce manual participation, this paper presents a novel and efficient feature extraction framework based on sparse autoencoder, which aims to extract high-dimensional and sparse features through two sparsity regularizers. Features obtained by the autoencoder can be exploited by different classifiers. Experimental results on the MIT-BIH database show that the classification performance of the proposed approach outperforms most of the state-of-the-arts.