PENGENALAN UCAPAN DENGAN JARINGAN SYARAF TIRUAN KUANTISASI VEKTOR ADAPTIF (SPEECH RECOGNITION WITH ARTIFICIAL NEURAL-NETWORK
ABSTRACT This research is aimed at supporting a system which is capable of recognizing speech produced by Indonesian speakers. The speech to recognize are the following phonemes: "a", "i", "u", "e", and "o". Training samples were taken from speakers...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Article NonPeerReviewed |
| 出版: |
[Yogyakarta] : Universitas Gadjah Mada
2001
|
| 主題: | |
| 在線閱讀: | https://repository.ugm.ac.id/24733/ http://i-lib.ugm.ac.id/jurnal/download.php?dataId=7711 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Universitas Gadjah Mada |
| 總結: | ABSTRACT
This research is aimed at supporting a system which is capable of recognizing speech produced by Indonesian speakers. The speech to recognize are the following phonemes: "a", "i", "u", "e", and "o". Training samples were taken from speakers of different sex, age, and ethnic groups (representing a variety of speakers). The method used for speech recognition was the artificial neural network, which is relatively new in pattern recognition.
The patterns of speech would be recognized after the feature extraction and classification were conducted. A neural network system would be able to recognize input pattern which had similarity with those given in a training process. During the training period the network carried out a feature extraction process followed by classification in which similar input patterns were classified into the same class. The trained network was then tested. For a given input vector the network would give an output vector according to the class of the input pattern.
The test results gave 79% fidelity for independent speakers speech recognition and 97% fidelity for single speaker speech recognition. The programs were written by MATLAB software.
Key words : speech, recognition, neural-network |
|---|