RNN BASED NOISE ANNOYANCE MEASUREMENT FOR URBAN NOISE EVALUATION

Noise is the unpleasant sound that may harm the activity or balance of human life. With the development of economic growth, urban noise pollution becomes a worldwide concern. Many efforts have been done in this area to evaluate, control, and reduce it. Among these efforts, noise-induced annoyance level evaluation attracts many attention which provides efficient tools for other noise control solutions. In this manuscript, an objective evaluation scheme is proposed which considers the time series effect and utilizes the Recurrent Neural Network (RNN) to generate noise annoyance score. The proposed solution does not require subjects to be involved for listening test which saves the time and can be implemented on-the-fly. Simulation results show that our proposed method presents satisfied prediction accuracy of noise annoyance level.