Electrocardiogram analysis method based on multi-kernel multi-scale convolutional neural network

The invention discloses an electrocardiogram analysis method based on a multi-kernel multi-scale convolutional neural network. The electrocardiogram analysis method comprises the following steps: S1, acquiring electrocardiogram waveform data; s2, constructing a multi-core multi-scale convolutional neural network algorithm model; s3, carrying out Db6 wavelet processing on the data set of the electrocardiogram waveform data by adopting a wavelet transformation mode; s4, performing data preprocessing on the electrocardiogram waveform data subjected to wavelet transformation in the step S3; s5, when the electrocardiogram waveform data is preprocessed, QRS waves need to be positioned; s6, obtaining an initial sample set, and dividing the initial sample set into a plurality of sample subsets according to different features; for different sample subsets, randomly extracting a plurality of samples to respectively generate a training set and a test set; and a k-fold cross validation experiment is adopted. Through test