1 / 14

A Tone Recognition Framework for Continuous Mandarin Speech

A Tone Recognition Framework for Continuous Mandarin Speech. Lei He, Jie Hao Toshiba (China) Research and Development Center INTERSPEECH 2006 - ICSLP. Hsiao- Tsung Hung. Introduction. LVCSR 結合聲調辨識 Embedded tone modeling: [MFCC + F0] Model the tone pattern separately. System Framework.

darryl-burns
Download Presentation

A Tone Recognition Framework for Continuous Mandarin Speech

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. A Tone Recognition Framework for Continuous Mandarin Speech Lei He, JieHao Toshiba (China) Research and Development Center INTERSPEECH 2006 - ICSLP Hsiao-Tsung Hung

  2. Introduction • LVCSR結合聲調辨識 • Embedded tone modeling: • [MFCC + F0] • Model the tone pattern separately

  3. System Framework

  4. F0 detection

  5. F0 detection • Normalized short-time autocorrelation function K. Hirose, H. Fujisaki, S. Seto, “A scheme for pitch extraction for speech using autocorrelation function with frame length proportional to the time lag”, Proc. ICASSP, Vol. I, pp. 149-152, 1992.

  6. Subsection outlined features

  7. Subsection outlined features E(F0)：average F0 value ： movement of F0 value E(VL)：average voicing level “the correlation coefficient of each frame is used to represent the voicing level .” *4 + duration = 13 (dimension) Base line

  8. Contextual Features Expansion Describe co-articulation effects *6 + duration = 13 (dimension)

  9. Contextual Tone Information

  10. Phonetic category information • All phonetic units are clustered into 7 classes according to corresponding phonetic attributes. • Using ID as features. • Add 5-dimension features: [pre-Final + Initial + Final + next-Initial + next-Final]

  11. Experiment

  12. Experiment

  13. Experiment

  14. Experiment

More Related