Speeding is an important factor influencing road traffic safety. Even though speed is monitored using radars, the drivers may increase speed after passing the radar. In this paper, we address automatic classification of speed changes (or maintaining constant speed) from audio data, as a microphone added to the radar can register the drivers’ behavior both in front of and behind the radar. We propose two time-frequency based approaches to represent the audio data for speed classification purposes. These approaches have been tested in a pilot study using on-road data, and the results are presented in this paper.