叠加去噪自动编码器结合深度神经网络的心电图信号分类方法

针对现有心电图信号分类方法精度较低,模型训练收敛速度较慢的缺点,提出一种基于叠加去噪自动编码器和深度神经网络方法的新型分类方法。该方法采用无监督学习方式,利用带有稀疏约束的叠加去噪自动编码器,实现心电图原始数据的特征学习。基于深度神经网络对信号进行分类,同时利用监督式自主学习微调方法对神经网络权重进行适时调整,从而保证信号分类的精度和质量。利用三个机构的经典数据库对该方法进行实验研究,并与目前两种最新的方法进行对比。实验结果证明,该方法在专家标记样本较少的情况下,仍能明显提高心电图数据分类的准确率,同时加快训练时的收敛速度。

ELECTROCARDIOGRAM SIGNALS CLASSIFICATION METHOD BASED ON STACKED DENOISING AUTOENCODER COMBINED WITH DEEP NEURAL NETWORK ALGORITHM

Facing with the weakness in the current traditional electrocardiogram( ECG) signals classification method, low precision and low model training convergence speed, we proposed a novel classification method based on stacked denoising autoencoder ( SDA ) and deep neural network ( DNN ). Under an unsupervised way, the SDA with sparsity constraint was adopted to learn suitable features from the raw ECG data. The DNN was utilized to classify ECG signals, and the supervised autonomic-learning fine-tuning method was used to regulate the DNN weights in real time to guarantee the speed and precision of the proposed method. Experiments on the classic database of three institutes were carried out to compare the proposed method with two state-of-the-art methods. The results indicate that, even with less expert- labelled samples, the proposed approach can still enhance accuracy for ECG classification significantly and accelerate the convergence speed in training.