5 ms·

ok, a compromise. let's do ML on the power spectral density of the train audio. or just use lidar.