Method for restoring the spectral components obtained by attenuating the test noise cancellation sound signal as a result of removing the noise in the test speech signal

摘要

It represents the training undistorted sound signal as a synthetic component of training unstrained basic components, by representing the training noise elimination audio signal as a synthetic component of training Yuibitsu basic components, test noise removal voice as a result of removing the noise of the test audio signal spectral components attenuated in the signal is restored. Test noise removal signal is decomposed as a synthetic component of training Yuibitsu basic component. After that, no-distortion test audio signal is estimated as a synthetic component of training unstrained basic component, the synthetic component is the same as the synthetic component of training Yuibitsu basic component.