Presentation is loading. Please wait.

Presentation is loading. Please wait.

Audio-visual Speaker association Zhijie Shao Master of Computer Science Supervisor: Trent Lewis.

Similar presentations


Presentation on theme: "Audio-visual Speaker association Zhijie Shao Master of Computer Science Supervisor: Trent Lewis."— Presentation transcript:

1 Audio-visual Speaker association Zhijie Shao Master of Computer Science Supervisor: Trent Lewis

2 Project Schedule Open MARYVoice Import ToolAusTalkA New VoiceEvaluation

3 Open MARY What is MARY? MARY – Modular Architecture for Research on speech sYnthesis Why is MARY?

4 Voice Import Tool Import new voices under the MARY environment. Two formats of files: 1.Wave files 2.Text files in MARY format Import Blizzard competition Data into MARY. (including English audiobook data and training data)

5 AusTalk AVSP requires large datasets. the largest-ever auditory-visual database of Australian speech HCSvLab Human Communication Science Virtual Laboratory A platform for eResearch in HCS

6 A New Voice Text Analysis Text Normalization Homonym Disambiguation Grapheme-to-Phoneme (Letter-to-Sound) Intonation Waveform Generation Unit Selection Diphones

7 Evaluation Emotion Quality intonation Consistent

8 Conclusion Final Object: Create a new voice Time schedule


Download ppt "Audio-visual Speaker association Zhijie Shao Master of Computer Science Supervisor: Trent Lewis."

Similar presentations


Ads by Google