I am currently working with audio from soccer video. Based on the belief that soccer video is filled with less eventful moments (i.e. goals, bookings, red cards, penalties etc. happen very rarely and are sparse).... I am assuming that if enough consecutive samples are taken, the resulting distribution will be Gaussian... where it will be dominated by lower audio measurements instead of high measurements (i.e. when events happen). So, is it ok to assume the data will be Gaussian in this case?