查詢結果分析
相關文獻
- 動態規正法應用於國語單音辨認的研究
- 國語單音的聲調辨認
- A Segmental Model Approach for Mandarin Final Recognition
- Tone Recognition for Continuous Mandarin Speech with Limited Training Data Using Selected Context-Dependent Hidden Markov Models
- 會聽國語的電腦
- Recognition of Mandarin Connected Digits
- 一種簡便的國語四聲辨認方法
- 各種連續語音辨認法則應用在國語數字上性能之比較
- A Comparative Study on the Performance of Several Speech Recognition Techniques Applied on the Highly Confusing Mandarin Syllables
- 類神經網路做國語數字辨認
頁籤選單縮合
| 題 名 | 動態規正法應用於國語單音辨認的研究 |
|---|---|
| 作 者 | 陳榮貴; 黃英峰; 劉繼謚; | 書刊名 | 電信研究 |
| 卷 期 | 21:1 1991.03[民80.03] |
| 頁 次 | 頁35-46 |
| 分類號 | 312.23 |
| 關鍵詞 | 動態規正法; 國語; 單音; 辨認; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 在語音辨認技術上,最重要的兩個考量因素是速度與辨認率。本文就動態規正法在國語音辨單認的應用上,探討各種演算構架與參數選取,並以實驗結果作一比較與分析。 在辦認速度方面,我們討論不同的搜尋方式以及將向量量化運用於動態規正比對法的效果,其中以向量量化為預處理的方式可提高速度至傳統比對法的三倍。 在辨認率方面,我們討論樣本訓練方式、每一字音參考樣本數量、倒頻譜係數階數與加權方式以及結合動態與瞬時語音參數的影響,其中結合動態與瞬時參數,並經正規化加權的辦認率為92.3%約比傳統方法提高4%。 |
| 英文摘要 | Speed and recognition rate are two of the most important factors considered in speech recognition. In the paper, we study the dynamic time warping method on Mandarin syllable recognition and discuss different kinds of algorithm and feature selecting. In recognition speed, we discuss the effect of searching algorithms and the application of Vector Quantization on DTW. The speed of VQ preprocessor method is 3 times as traditional DTW method. In recognition rate, we discuss the number and training method of reference templates, the order and weighting of cepstrum coefficient, and the combination of temporal and dynamic speech features. In the case of combined features and weighted by normalizing method, the recognition accuracy is 92.3%, which is 4% higher than traditional method. |
本系統中英文摘要資訊取自各篇刊載內容。