1997 Fiscal Year Final Research Report Summary
A Recognition System of Japanese and Korean for the Robot
| Project/Area Number |
08045045
|
|---|
| Research Category |
Grant-in-Aid for international Scientific Research
|
| Allocation Type | Single-year Grants |
|---|
| Section | University-to-University Cooperative Research |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Miyagi national College of Technology |
|---|
Principal Investigator |
KARASAWA Shinji Miyagi national tech., Electric Eng., Professor, 電気工学科, 教授 (10042243)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
鄭 俊憲 仁荷工業専門大学, 電子科, 教授
具 龍會 仁荷工業専門大学, 電子科, 教授
IKEDA Senri Miyagi national tech., Materials Eng., Professor, 材料工学科, 教授 (60109832)
CHUNG Jun Hum Inha tech., Electronics, Professor
KU Yang Hea Inha tech., Electronics, Professor
|
|---|
| Project Period (FY) |
1996 – 1997
|
| Keywords | the Japanese language / the Korean language / waveforms of speech / speech coding / speech understanding / language processing / telephone interpretation |
|---|
| Research Abstract |
Speech processing and language processing are investigated and examined in the research project entitled "A recognition system of Japanese and Korean for the robot", the results of the investigation in this stage give fundations for spoken language system such as "Automatic Telephone Interpretation between Japanese and Korean".
One of the result entitled "Language Processing which is Carried out by Decoders" was written by Japanes characters and it was translated into Korean written by "Hangil" characters. the interpretation shows the similarities between the Japanese language and the Korean language. But it also shows the differences in phonations. The phonetion is represented by each character i.e. "kana" and "Hangil".
Electronic circuits are made in order to distinguish Japanese vowels one-from the other. The parallel data in a pich of vowel assign the phonetic symbol by mean of decoders. Since circuits for recognition of consonants are complicated, the computational model known as "cepstral method" in speech analysis is studied. The method uses a codebook for determination of phonetic symbol where cepstral distances can be calculated from parameters on linear predictive coefficients and the stream of phonetic symbols can be transfored to words by the parallel decoders.
|
