Structure optimization of metamodels to improve speech recognition accuracy

The metamodels is a technique that was developed to model a speaker's phoneme confusion-matrix and use this information to increase speech recognition accuracy for speakers with disordered and normal speech. Approaches to improve the performance of the metamodels, mainly focused on obtaining better estimates of the speaker's confusion-matrix, were studied. While some achieved significant improvements, alternatives to the functional structure of the metamodels were not explored. In this paper is proposed a different structure for the metamodel of a phoneme and its optimization by means of a genetic algorithm. Results showed statistically significant gains in speech recognition accuracy over the previous metamodels.