eduzhai > Applied Sciences > Engineering >

A New Approach to Accent Recognition and Conversion for Mandarin Chinese

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: Two new approaches to accent classification and conversion are presented andexplored, respectively. The first topic is Chinese accentclassification recognition. The second topic is the use of encoder-decodermodels for end-to-end Chinese accent conversion, where the classifier in thefirst topic is used for the training of the accent converter encoder-decodermodel. Experiments using different features and model are performed for accentrecognition. These features include MFCCs and spectrograms. The classifiermodels were TDNN and 1D-CNN. On the MAGICDATA dataset with 5 classes ofaccents, the TDNN classifier trained on MFCC features achieved a test accuracyof 54 and a test F1 score of 0.54 while the 1D-CNN classifier trained onspectrograms achieve a test accuracy of 62 and a test F1 score of 0.62. Aprototype of an end-to-end accent converter model is also presented. Theconverter model comprises of an encoder and a decoder. The encoder modelconverts an accented input into an accent-neutral form. The decoder modelconverts an accent-neutral form to an accented form with the specified accentassigned by the input accent label. The converter prototype preserves the toneand foregoes the details in the output audio. An encoder-decoder structuredemonstrates the potential of being an effective accent converter. A proposalfor future improvements is also presented to address the issue of lost detailsin the decoder output.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×