I just want to add that PeakUtils also support fitting gaussians and computing centroids to increase the peak resolution, allowing for a higher resolution (instead of just finding the indexes). Also, the PeakUtils indexes function was inspired on Marcos Duarte implementation, so those methods have some similarities in the results.
I had to calculate (well, estimate) the SNR of signals for which I had little useful information. So, I needed to isolate the 'signal' (my signals had only one peak) from the rest of the waveform to get an estimate for the noise. Bah, not fun.