Neuro-fuzzy estimator of speech quality

Abstract

A speech quality estimator based on neuro-fuzzy techniques is presented in this paper. The proposed estimator employed a first-order Sugeno type fuzzy inference system (FIS) to estimate speech quality only using the output signal of a system under test. The features utilized by the proposed estimator were derived from the perceptual spectral density of input speech. The premise and consequent parameters of the FIS were constructed by an adaptive neuro-fuzzy inference system (ANFIS) and tuned by the back-propagation and least squares algorithms. The performance of the proposed estimator was demonstrated using speech codec and pathological voice data sets.