Abstract: As an important indicator to evaluate the influences of noises on human, annoyance has been a hot topic in the field of acoustics and environmental science. The annoyance of traffic noise is modeled from the perspective of timbre attributes, and the physical interpretation of annoyance dimensions is given. The results indicate that (1) the timbre space of traffic noise annoyance is consisted of four dimensions, (2) the obtained model for annoyance computation based on linear regression approach is good agreement with the sound samples, and (3) the annoyance dimensionality correlates well with certain spectral characteristics of the noise signals. The results offer suggestions on annoyance modeling and noise control.