A detection algorithm aimed at analysing localized intermittent oscillations in sea level time series and identifying meteotsunami events is presented. The algorithm is based on wavelet analysis of high frequency sea level oscillations recorded by tide gauges. Detection thresholds are fixed both for oscillation amplitude and averaged power spectral density for each tide-gauge station. This method is tested on different historic meteotsunami events and during the large event of 26 July 2012. This method is also used to characterize harbour seiche oscillation periods in the North-Western Mediterranean Sea.