This uses findpeaks to locate the impuse signal in channel 2 of the wave file. Peaks are required to be at least 3 standard deviations larger than the mean signal. The values are replaced with the median of that channel (excluding the impulse signal). fFilt is the filtered version of f.
% Read in wav file
FileName = 'rec_noise.wav'; %full path is better
[f,fs] = audioread(FileName);
% Locate peaks
minHeight = mean(abs(f(:,2))) + std(abs(f(:,2)))*3; % min height to be 'peak'; 3 std above mean
[pks, locs] = findpeaks(abs(f(:,2)), 'MinPeakHeight',minHeight);
pks = pks .* sign(f(locs,2)); % pks is only used for plotting
% Replace peaks with the median value of f(:,2) expluding outliers.
% You could also just remove those values but you'd need to remove them in chan 1, too.
isOut = ismember(1:size(f,1), locs);
replacementValue = median(f(~isOut,2)); % you could also use NaN, 0, etc....
fFilt = f;
fFilt(isOut,2) = replacementValue;
Plot the results.
clf()
% Plot channel 1
sp(1) = subplot(3,1,1);
plot(sp(1), f(:,1),'DisplayName','chan 1')
title(sp(1), 'channel 1')
% Plot channel 2
sp(2) = subplot(3,1,2);
plot(sp(2), f(:,2),'DisplayName','chan 2')
title(sp(2), 'channel 2')
% Show outliers (impulses)
hold(sp(2),'on')
plot(sp(2),locs, pks, 'rx', 'DisplayName', 'Outliers')
% Plot the filtered channel 2
sp(3) = subplot(3,1,3);
plot(sp(3), fFilt(:,2),'DisplayName','chan 2 Filtered')
title(sp(3), 'channel 2 filtered')
linkaxes(sp)
