頁籤選單縮合
| 題 名 | 碼字激發線性預估編碼之隨機碼本的改進研究=A Study on Modified CELP Stochastic Codebook of Speech Coding |
|---|---|
| 作 者 | 廖鴻儒; | 書刊名 | 中華工商專校學報 |
| 卷 期 | 15 1997.09[民86.09] |
| 頁 次 | 頁21-27 |
| 分類號 | 312.23 |
| 關鍵詞 | 隨機碼本; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | F1016 CELP隨機碼本(Stochastic Codebook)屬三位階碼本(-1,0,+1)之設計 ,雖可降低計算複雜度,但在激發源之資訊提供上又嫌不足,造成合成音在聽覺上,低音的 清析度較為不佳,有沙啞聲出現。因為在次佳化搜尋過程中,原音減去調適碼本中之最佳碼 字的合成音後, 仍保有類似 Pitch (高音)之資訊,而使用的三位階隨機碼本所提供之激 發源資訊,仍無法提供完整的補償。因此在不增加過多的計算量之條件下,在三位階隨機碼 本中, 加入適當的脈衝信號, 用來補償不足的資訊部分, 經模擬實驗證明, 對於改善 SegSNR (聽覺評估客觀指標)及主觀聽覺效果,皆有不錯的表現。 |
| 英文摘要 | The CELP stochastic codebook have simple values of three-level. This architecture can simplify complexity, but it doesn't provide sufficient excited information. So synthetic speech quality become bad and appear hoarse sounds in low-bit rate. In otherwise, the residual signals of CELP still have similar pitch in the waveforms. It indicates that speech coding of CELP have lost a part information for parameter collections. In this reason, three-level codebook increases several impulses to compensate for a part excited information. In the simulated result, we improve SegSNR with raised values and subjective percept as well. |
本系統中英文摘要資訊取自各篇刊載內容。