A couple of points:
1) Taking the average of the channels is not usually recommended for mixing down to stereo. Instead, it is common to sum the two channels but normalize them according to the absolute value of the maximum peak in either channel (https://en.wikipedia.org/wiki/Audio_normalization). 2) Hemanji appears to be asking to conditionally mix the stereo down. That is, the sound file may already be mono in which case we would get an error if we try to mix it down.
An example script that accomplishes all of this might look like this:
[x,fs] = audioread('myWavFile.wav');
[m, n] = size(x); %gives dimensions of array where n is the number of stereo channels
if n == 2
y = x(:, 1) + x(:, 2); %sum(y, 2) also accomplishes this
peakAmp = max(abs(y));
y = y/peakAmp;
% check the L/R channels for orig. peak Amplitudes
peakL = max(abs(x(:, 1)));
peakR = max(abs(x(:, 2)));
maxPeak = max([peakL peakR]);
%apply x's original peak amplitude to the normalized mono mixdown
y = y*maxPeak;
else
y = x; %it is stereo so we will return it as is (e.g., for additional processing)
end
