Heart Rhythm Abnormal Signal Diagnosis Based on Neural Network Deep Learning

The main endpoint drift, electromyography (EMG) interference signals, step-up transformer interference signals, and large motion artifacts often appear in ambulatory rhythm. In solving the signal problem, the traditional method has caused a great loss. The deep learning neural network model used in this study did not require prior knowledge related to the characteristic waveforms and pathological features. Using supervised or unsupervised learning of various features related to the data and classification, the limitations caused by insufficient prior knowledge were avoided. We proposed the form of pre-reinforcement training with the model. Using a deep neural network, the unsupervised learning of data for ECG examination was achieved. By pre-training and manually adjusting the experimental comparison of multiple databases, the calculation accuracy of the model was effectively improved. The information associated with the extrinsic features of the extracted data was adopted for learning reinforcement training. The fusion of the control mechanisms enhanced the received signal containing the generated noise and contributed to the extraction of useful extrinsic features.