- 期刊
基於隱藏式條件隨機域聲學模型之鑑別式訓練法研究
An Investigation of Acoustic Modeling Techniques with Hidden Conditional Random Field
摘要
本文探討採用隱藏式條件隨機域（Hidden Conditional Random Field；簡稱HCRF）於語音辨識之聲學模型，並與傳統之隱藏式馬可夫模型（Hidden Markov Model；簡稱HMM）進行分析比較，並提出一個結合鑑別式法則之新穎HCRF模型訓練方法。經由TEST500語音資料庫進行連續音節辨識的實驗結果，發現HCRF有較佳的辨識率，其辨認反應時間遠快於HMM，更適合運用於即時辨識。我們利用鑑別式法則訓練HMM至收斂，將其參數轉換成HCRF初始參數，並繼續使用鑑別式法則訓練HCRF模型，其效能相較於最大相似度法則訓練出的HMM，提高了10.7%相對音節正確率。
並列摘要
In this paper, we adopt an acoustic modeling with Hidden Conditional Random Field (HCRF)-based approach for speech recognition; and its performance is compared with the traditional Hidden Markov Model (HMM) in the same structure. A novel HCRF training algorithm combining the discriminative training criterion is proposed. As shown in experimental results of the continuous Mandarin syllable recognition in TEST500 database, the HCRF-based approach performs better than the one obtained with HMM in the accuracy rate and response time. Proved by a serial of related experiments, HCRF is more suitable for real-time speech recognition system.
延伸閱讀
- 羅天宏、陳柏琳（2018）。結合鑑別式訓練與模型合併於半監督式語音辨識之研究。中文計算語言學期刊，23(2)，19-34。https://www.airitilibrary.com/Article/Detail?DocID=1027376x-201812-202001030015-202001030015-19-34
- 邢凱婷（2009）。基於隱藏式條件隨機域語者模型之語者識別演算法〔碩士論文，元智大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0009-2807200914245700
- 羅天宏（2019）。探討聲學模型化技術與半監督鑑別式訓練於語音辨識之研究〔碩士論文，國立臺灣師範大學〕。華藝線上圖書館。https://doi.org/10.6345/THE.NTNU.DCSIE.004.2019.B02
- Leung, M. Y. (1983). Probabilistic modelling of some problems in computer science [master's thesis, The University of Hong Kong]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0029-1812201200005657
- Billon, A., Picaut, J., Valeau, V., & Sakout, A. (2012). Acoustic Predictions in Industrial Spaces Using a Diffusion Model. Advances in Acoustics and Vibration, 2012(), 119-127. https://doi.org/10.1155/2012/260394