Do you have a particular selection criteria for the peaks or formula for how you would like the smoothing performed at the extremities?
I ran a spectrogram on your signal to see if there was a simple envelope to extract. It appears that there is, so maybe you could try to extract the amplitude ridges. (It seems a little bit of aliasing is in your sampled signal as well. Not sure if it's important, but have a look)
spectrogram(y,kaiser(256,40),250,[],1/mean(diff(x)),'power','yaxis')
I tried to extract a few ridges:
[sst,f] = fsst(y,1/mean(diff(x)),kaiser(256,10));
fridge = tfridge(sst,f,10,'NumRidges',6);
plot(x,fridge)
I got a reasonable hit on the fundamental, so perhaps you could attempt to extract the spectrum amplitudes along those frequencies to get a good sense of the power envelope.
