Fix time-domain calculations of the optimal observation filter

The calculations for the time-domain optimal observation filters have been fixed.

There are some minor improvements in other fields such as:

• A bug in obsFiltTD() that screwed up the dimensions of the arrays when there was only one source.
• A bug in ptSrcFieldTD() that screwed up the dimensions of the arrays when there was only one source.
• The virtual microphone geometry generation function was heavily modified, and its interface has changed.

This merge closes the Closes #8 (closed).

Virtual Sensing/Remote Microphone Technique/MATLAB/Functions/obsFiltTD.m

@@ -211,27 +211,24 @@ function [O, Rme, Rmm, Ovec, RmeMtx, RmmMtx, condNum, mMtx, Omean, RmeMean, RmmM
         for mIdx = size(m, 2):-1:1
             % Calculate the cross-correlations between virtual and monitoring microphones
             for eIdx = size(e, 2):-1:1
-                [corr, lags] = xcorr(m(:, mIdx, jIdx), e(:, eIdx, jIdx));
-                lIdx = find(lags == 0);
+                corr = xcorr(m(:, mIdx, jIdx), e(:, eIdx, jIdx), filtLen, "unbiased");
 
-                Rme(:, mIdx, eIdx, jIdx) = corr(lIdx:-1:lIdx-filtLen + 1);
+                Rme(:, mIdx, eIdx, jIdx) = corr(filtLen + 1:-1:2);
             end
 
             % Go through the monitoring microphones to calculate the monitoring microphone correlation matrices
             for mmIdx = mIdx:-1:1
                 % Auto-correlation matrices are Toeplitz symmetric
                 if mIdx == mmIdx
-                    [corr, lags] = xcorr(m(:, mmIdx, jIdx), m(:, mmIdx, jIdx));
-                    lIdx = find(lags == 0);
+                    corr = xcorr(m(:, mmIdx, jIdx), m(:, mmIdx, jIdx), filtLen, "unbiased");
 
-                    Rmm(:, :, mIdx, mmIdx, jIdx) = toeplitz(corr(lIdx:-1:lIdx - filtLen + 1));
+                    Rmm(:, :, mIdx, mmIdx, jIdx) = toeplitz(corr(filtLen + 1:-1:2));
                 else
-                    [corr, lags] = xcorr(m(:, mIdx, jIdx), m(:, mmIdx, jIdx));
-                    lIdx = find(lags == 0);
+                    corr = xcorr(m(:, mIdx, jIdx), m(:, mmIdx, jIdx), filtLen, "unbiased");
 
                     % Cross-correlation matrices
                     for iIdx = filtLen-1:-1:0
-                        Rmm(:, iIdx + 1, mIdx, mmIdx, jIdx) = corr(iIdx + (lIdx:-1:lIdx - filtLen + 1));
+                        Rmm(:, iIdx + 1, mIdx, mmIdx, jIdx) = corr(iIdx + (filtLen + 1:-1:2));
                     end
                     Rmm(:, :, mmIdx, mIdx, jIdx) = squeeze(Rmm(:, :, mIdx, mmIdx, jIdx)).';
                 end
@@ -243,8 +240,8 @@ function [O, Rme, Rmm, Ovec, RmeMtx, RmmMtx, condNum, mMtx, Omean, RmeMean, RmmM
     % Post-process cross- and auto-correlation matrices
     % ====================================================
     % "Reshape" the data
-    RmeMtx = reshape(Rme, prod(size(Rme, [1, 2])), size(Rme, 3), size(Rme, 4));
-    RmmMtx = reshape(permute(Rmm, [1, 3, 2, 4, 5]), prod(size(Rmm, [1, 3])), prod(size(Rmm, [2, 4])), size(Rmm, 5));
+    RmeMtx = reshape(permute(Rme, [2, 1, 3, 4]), prod(size(Rme, [1, 2])), size(Rme, 3), size(Rme, 4));
+    RmmMtx = reshape(permute(Rmm, [3, 1, 4, 2, 5]), prod(size(Rmm, [1, 3])), prod(size(Rmm, [2, 4])), size(Rmm, 5));
 
     % ====================================================
     % Calculate observation filters