Algorithm 1: Preprocessing the input signal using Discrete Wavelet Transform algorithm

Input: original noisy speech signal, wavelet decomposition bands Output: decomposed signals and corresponding coefficients

Xdata[] stores the input data vector, and Ydata[ ] is the output data vector that is returned. N is the length of both data vectors. Before applying this approach, it is presumable that the wavelet filter parameters G[k] and the scale filter parameters H[k] have been provided. L is the total number of parameters. N must be an even number to work with this algorithm.

Step 1: Set $\mathit{s}=\frac{\mathit{N}}{2}$ // Start index of the input array’s gamma coefficients

Step 2: Allocate $\mathit{ydata}\left[\mathit{N}\right];$ // Provide a memory space for the output data vector

Step 3: for (i = 0 while i <  N increment i  =  i + 1) do // loop over input data

Step 4:$\mathit{ydata}\left[\mathit{i}\right]=0;$ // Reset summation accumulators.

Step 5: endfor;

Step 6:j = 0; // access/index to the output data array

Step 7: for (i = 0 while i <  N increment i  =  i + 1) do // loop over input data.

Step 8: for k = 0 while k < L increment k = k + 1) do // convolution loop.

Step 9:$\mathit{didx}=\left(\mathit{i}+\mathit{k}\right)\mathit{mod}\mathit{N}$; // access/index into input data with wraparound.

Step 10:$\mathit{ydata}\left[\mathit{j}\right]=\mathit{ydata}\left[\mathit{i}\right]+\mathit{G}\left[\mathit{k}\right]\ast \mathit{xddata}\left[\mathit{didx}\right]$; // Scaling filter contribution

Step 11:$\mathit{ydata}\left[\mathit{j}+\mathit{s}\right]=\mathit{ydata}\left[\mathit{i}+\mathit{s}\right]+\mathit{H}\left[\mathit{k}\right]\ast \mathit{xddata}\left[\mathit{didx}\right];$ // Wavelet filter contribution

Step 12: endfor;

Step 13:j = j + 1;  // Update position in output array

Step 14: endfor;