Algorithm 2. Noise analysis and reduction using OCNMF

Input: Spectrogram of an input signal $\mathit{Y}$, priori basis number R, an undesired basis number J, type of β-divergence β,

number of iterations in OCNMF $i_{\mathit{OCNMF}}$, shift length of CNMF K, signal basis matrix ${\mathit{H}}_{\tau }$,

weight parameter $\mu$, phase matrix of the input signal ${\mathit{Y}}_p$

Output: Noise reduced signal $Y_{\mathit{cl}}$

1:  Initialize ${\mathit{F}}_{\tau }$, U, and G with random non-negative values

2:  Normalize columns of ${\mathit{F}}_{\tau }$

3:  for i = 1, $\cdots$, $i_{\mathit{OCNMF}}$ do

4:   for $\tau$ = 0, $\cdots$, R − 1 do

5:    Compute Z

6:    Update ${\mathit{F}}_{\tau }$, U, and G using Equations (15), (16) and (23)

7:    Normalize columns of ${\mathit{F}}_{\tau }$

8:   end for

9:  end for

10:  Compute M using Equation (19) and ${\mathit{Y}}_{\mathit{cl}}$ = $\mathit{Y}$ $⨀$ M

11:  ISTFT (${\mathit{Y}}_{\mathit{cl}}$ $⨀$ ${\mathit{Y}}_p$)