Assessment of speech intelligibility in scholar classrooms by measurements and prediction methods

According to Italian regulation, the Ministerial Decree of 11 October 2017 about Environmental Criteria, reference values for acoustic indoor quality descriptors in public buildings are imposed. Regarding school environments, indoor acoustic quality targets refer to reverberation time, clarity, and speech intelligibility, whose representative acoustic descriptor is the speech transmis- sion index (STI). This paper presents pyeSTImate, a Python-based tool for speech transmission index prediction in lecture rooms. The tool returns fully simulated results from the dimensions and material characteristics of classrooms with parallelepiped geometry and without limitations in size. Extensive experiments have been conducted with different simulation methods, evaluating the accuracy by comparison with in situ measurements selected from primary, secondary, and university classrooms in school buildings of the Marche Region in Italy. The combination of simu- lated speech transmission indexes with a prediction method based on an artificial neural network has also been evaluated. The analysis of the performance demonstrates the computational robust- ness of the tool that enables its use for the analysis of existing rooms, as well as for the renova- tion and design of new spaces.