ClassTalk system for predicting and visualizing speech in noise in classrooms

This paper discusses the ClassTalk system for modeling, predicting and visualizing speech in noise in classrooms. Modeling involves defining the classroom geometry, sources, sound-absorbing features, and receiver positions. Empirical models, used to predict speech and noise levels, and reverberation times, are described. Male or female speech sources, and overhead-, slide-, or LCD-projector, or ventilation-outlet noise sources, can have four output levels; values are assigned based on ranges of values found from published data and measurements. ClassTalk visualizes the floor plan, speech- and noise-source positions, and the receiver position. The user can walk through the room at will. In real time, six quantities—background-noise level, speech level, signal-to-noise level difference, useful-to-detrimental energy fraction (U50), Speech Transmission Index, and speech intelligibility—are displayed, along with occupied and unoccupied reverberation times. An example of a large classroom before and after treatment is presented. The future development of improved prediction models and of the sound module, which will auralize speech in noise with reverberation, is discussed.