Confirmation Research on Robustness Speaker Based-on Normalized Autocorrelation Voice Selection
This paper gives a new algorithm that based on normalized autocorrelation robustness selection. Aiming at the situation that the voice was vitiated by noise and the difference of pollution extent between different parts of the voice, we recognized that preserving the polluted parts of voice will bring in negative affection to the confirmation system of speaker. Normalized autocorrelation not only shows the correlation of short-time voice frame, but also shows the polluted extent of voice. As a result, we can wipe off the deeply polluted voice through the normalized autocorrelation. Experiments have shown that through the procedure of normalized autocorrelation we can not only wipe off the polluted voice, but also improve confirmation ability of speaker in high noise environment.
C. Gu and S. Wang, "Confirmation Research on Robustness Speaker Based-on Normalized Autocorrelation Voice Selection", Applied Mechanics and Materials, Vols. 130-134, pp. 68-71, 2012