Speech synthesis by interpolation of prestored filter parameters
| Autor: | H. B. Ritea, R. A. Gillmann |
|---|---|
| Rok vydání: | 1977 |
| Předmět: |
Acoustics and Ultrasonics
Series (mathematics) business.industry Computer science Acoustics Linear prediction Speech synthesis Pattern recognition computer.software_genre Sample (graphics) Set (abstract data type) Filter design Arts and Humanities (miscellaneous) Filter (video) Artificial intelligence business computer Interpolation |
| Zdroj: | The Journal of the Acoustical Society of America. 62:S62-S62 |
| ISSN: | 0001-4966 |
| DOI: | 10.1121/1.2016291 |
| Popis: | A series of experiments as performed to compare the relative subjective quality of synthetic speech produced by several different sets of prestored filter parameters. These include (1) linear prediction coefficients, (2) reflection coefficients (k parameters), and (3) vocal‐tract area functions. For each experiment, a set of filter coefficients for both phoneme nuclei and transitions were extracted from prestored and preprocessed digitized speech. Continuous sets of filter coefficients were generated by interpolation. Both linear and higher‐order interpolation schemes were tested. Synthetic speech was also generated using only phoneme nuclei, in order to quantify the contribution of the transitional segments to overall speech quality. The various synthesis methods were subjectively evaluated by a panel of listeners. The results of this evaluation, along with sample recordings, will be presented. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|