eduzhai > Applied Sciences > Engineering >

Deep MOS Predictor for Synthetic Speech Using Cluster-Based Modeling

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: While deep learning has made impressive progress in speech synthesis andvoice conversion, the assessment of the synthesized speech is still carried outby human participants. Several recent papers have proposed deep-learning-basedassessment models and shown the potential to automate the speech qualityassessment. To improve the previously proposed assessment model, MOSNet, wepropose three models using cluster-based modeling methods: using a globalquality token (GQT) layer, using an Encoding Layer, and using both of them. Weperform experiments using the evaluation results of the Voice ConversionChallenge 2018 to predict the mean opinion score of synthesized speech andsimilarity score between synthesized speech and reference speech. The resultsshow that the GQT layer helps to predict human assessment better byautomatically learning the useful quality tokens for the task and that theEncoding Layer helps to utilize frame-level scores more precisely.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×