Algorithm 1: The proposed COBLS

Inputs: heartbeat matrix $X$ and heartbeat label $Y$

Outputs: network weight $W_n^m$

1: for ICA convolution do

2:  Calculate the whitening matrix ($V_L$) using Equation (8).

3:  Calculate the unmixing matrix ($D_L$) using the FastICA toolkit.

4:  Calculate the ICA convolutional kernel ($W_l^1$) using Equation (9).

5:  The initial eigenmatrix (${\mathrm{\Gamma }}_i^l$) is calculated using Equation (10).

6: end for

7:  for PCA convolution do

8:  Construct the second-order pending matrix ($U$).

9:  Calculate the covariance matrix ($U{U}^T$).

10:   Extract the co-eigenvector.

11:   Calculate the PCA convolutional kernel ($W_{\zeta }^2$) using Equation (11).

12:   Calculate the second-order eigenmatrix ($O_i^l$) using Equation (12).

13: end for

14: Calculate the decimal matrix ($T_i$) using Equation (13).

15: Calculate the histogram feature vector ($f_i=\left[Bhist\left(T_i^1\right),Bhist\left(T_i^2\right),\cdot \cdot \cdot ,Bhist\left(T_i^p\right),\right]$).

16: Take $f$ as the feature node ($Z^n$) of the network.

17: for j = 1:m do

18:  Generate random matrices $W_{h_j}$ and ${\beta }_{h_j}$.

19:  Calculate the enhancement node ($H_j$) using Equation (2).

20: end for

21: Calculate the network weight ($W_n^m$) using Equation (4).