Listening Difficulty Rating Meter Using Machine Learning for Assessing Public-Address Systems

摘要

Subjective speech quality assessment has been used widely for the development of outdoor public-address (PA) systems; however, this assessment has some difficulties in many cases. Therefore, we propose an objective listening difficulty rating meter for PA systems, which is based on the subjective listening difficulty rating prediction model, using the random forest algorithm and mel-frequency cepstrum coefficients. The performance of the proposed meter shows a high correlation (0.88) with the subjective evaluation results.