Prediction of psychoacoustic parameters using radial basic functions neural network

Abstract

The ability of human listeners to estimate psychoacoustic parameters from the speakers speech levels at different listener's positions in a classroom is an interesting and not yet thoroughly examined phenomenon. The goal of this paper is to propose a technique that uses radial basis functions neural network to estimate the non linear mapping function that best represents the relationship among input (Frequency Spectrum of Speech level at all listeners positions in the classroom, the sound pressure level and the signal to noise ratio of the speech signal) and output (Psychoacoustic parameters such as loudness, sharpness and roughness) variables in a database.