AN EFFICIENT FEATURE MATRIX FOR URBAN NOISE ANNOYANCE MEASUREMENT

Traditionally the noise measurement and monitoring are based on the averaged A-weighting sound pressure level. However it is not adequate to use this feature only to evaluate sound induced annoyance as human noise tolerance is subjective to sound type and its characteristics. In this paper, we propose an efficient noise annoyance measurement methodology for online/onsite evaluation. We use a set of sound features and analyze their influence to the annoyance using forward iterative method. We develop an algorithm for feature significance evaluation with correlation indicator as benchmark tool. The analysis results show that the 90 percentile instantaneous loudness and the minimum value of the 3rd component of tonality are the two most significant features for annoyance measurement. The new two dimensional feature matrix effectively reduce the computational complexity for urban noise annoyance level evaluation. The effectiveness of proposed annoyance matrix is verified using regression methods.