Automatic arrhythmia analysis method based on lead deep neural network

摘要: The invention discloses an automatic arrhythmia analysis method based on a lead depth neural network. includes generating multichannel electrocardiogram sample by three sampling methods;splicing obtained 600-dimensional signals along the second dimension; forming six 600*1-dimensional signal samples equivalently when original electrocardiogramsignals are in two leads; inputting into plurality of channels parallel respectively, wherein each channel is composed convolution layer units connected series, combination at output end combines characteristic patterns all last dimension, and attention located between LSTM unit. Each unit comprises extracting characteristics one-dimensional using excitation operation pooling series. series with all-connection as softmax. further outputting; learning parameters deep network automatically identifying sample. solves problem that existing system not enough to meetthe accuracy requirements clinical applications.