Algorithm 1: CL-RNMF.

Require: $x\left(t\right)$, $K_r$, $K_w$, $\alpha$, $\beta$, $\lambda$, and M.ss

1:  Compute the normalized magnitude spectrogram $\mathbf{X}$ using Equation (11).ss

2:  Initialize ${\mathbf{B}}_W$, ${\mathbf{B}}_R$, ${\mathbf{A}}_W$, and ${\mathbf{A}}_R$ with random non-negative values.ss

3:  Update the estimated wheezing basis matrix ${\mathbf{B}}_W$ using Equation (14).ss

4:  Update the estimated respiratory basis matrix ${\mathbf{B}}_R$ using Equation (15).ss

5:  Update the estimated wheezing activations matrix ${\mathbf{A}}_W$ using Equation (16).ss

6:  Update the estimated respiratory activations matrix ${\mathbf{A}}_R$ using Equation (17).ss

7:  Repeat Steps 3–6 until the algorithm converges (or until the maximum number of iterations M is reached).ss

8:  Compute the spectral energy distribution $\xi \left(f\right)$ from ${\mathbf{B}}_W$ using Equation (18).ss

return $\xi \left(f\right)$