Automatic evaluation of noise suppression in speech signal recorded during phonation in the open-air MRI

The paper is focused on the evaluation of successfulness of noise reduction of the speech signal recorded in an open-air magnetic resonance imager during a phonation for the 3D human vocal tract modeling. In more detail is there described the automatic evaluation method based on Gaussian mixture models (GMM) classification. Performed first-step experiments have successfully confirmed that the proposed GMM classifier of the speech quality is functional and fully comparable with the standard evaluation based on the listening test.