The identification of a special kind of acoustic events such as dual gunshots and single gunshots in the traffic background is described in this work. The recognition of dangerous sounds may help to prevent the abnormal or criminal activities that happened near to the public transport stations. Therefore in this paper the methodology of dual shots detection in a noisy background was developed and evaluated. For this purpose, we investigated various feature extraction methods and combinations of different feature sets. These approaches were evaluated by the widely used classification technique based on the Hidden Markov Models.