Spectral features for the classification of civilian vehicles using acoustic sensors

Identification of civilian vehicles is in increasing demand in military and homeland security applications such as surveillance and traffic monitoring. This paper presents a new algorithm to estimate the fundamental frequency of a car engine sound accurately. It also presents examples of the use of this frequency to identify and classify different civilian vehicles. Acoustic features, such as the tristimulus response introduced in music theory, are proposed as tools for the characterization of the timbre of the car engine sound. In addition we introduce a new generalized M-stimulus as a tool to distinguish special features in some car engines.