Investigation of speech intelligibility using artificial neural network model

Abstract EN

A classroom acoustic is an important and difficult part of university classroom design.

Good design is achieved more on the basis of acoustics expertise than on pure engineering design.

In this paper, the Artificial Neural Network (ANN) model is used for predicting speech intelligibility in classroom.

There are several classroom properties such as diminution of the class, signal to noise ratio (SNR), the location of the student and teacher , background noise where collected from the classroom.

A set of word is complied and a speech signal data base was created.

The sound pressure levels are then measured using sound pressure meter at different classroom positions.

A datasheet was obtained from the measurement and then used to provide as training database into learning process of (ANN) to predict the speech intelligibility at various listeners' position of classroom.

This method improve high accuracy, efficiency and economic of calculation intelligibility in classrooms.

Therefore it reduces the error by using the classic methods.