Prediction of Reverberation time in university classrooms using Neural Network

Abstract

Reverberation time is fundamental to the study of the acoustics of an enclosed space. An important objective of architectural acoustics is to predict the reverberation time in an enclosed space. Reverberation time is also very important in our daily life for a better hearing and communication. Using existing computer models, reverberation time predictions are too difficult and too inaccurate. This paper presents a method for predicting the reverberation time in university classrooms using neural network. Measurement of reverberation time was conducted in 8 lecture halls at University Malaysia Perlis. All the measurements were taken with furniture and without furniture. The quality of the sound in any enclosed lecture hall depends on the shape of the room, the furniture, floor area and other ceiling materials. Neural network is trained using Conventional Back Propagation (BP) algorithm.