1 Technical University of Denmark2 Technical University of Denmark
An artificial neural network structure has been specified, implemented and optimized for the purpose of predicting the perceived sound quality for normal-hearing and hearing-impaired subjects. The network was implemented by means of commercially available software and optimized to predict results obtained in subjective sound quality rating experiments based on input data from an auditory model. Various types of input data and data representations from the auditory model were used as input data for the chosen network structure, which was a three-layer perceptron. This network was trained by means of a standard backpropagation procedure and tested on selected stimuli from the subjective rating experiment. The best results were obtained with an additional input to the network, identifying the listener, and thus allowing different states for each subject. The performance with previously unseen test was evaluated for two types of test set extracted from the complete data set. With a test set consisting of mixed stimuli, the prediction error was only slightly larger than the statistical error in the training data itself. Using a particular group of stimuli for the test set, there was a systematic prediction error on the test set. The overall concept proved functional, but further testing with data obtained from a new rating experiment is necessary to better assess the utility of this measure. The weights in the trained neural networks were analyzed to qualitatively interpret the relation between the physical signal parameters and the subjectively perceived sound quality. No simple objective-subjective relationship was evident from this analysis.