以長短期記憶模型進行筆寫單字辨識｜國立清華大學博碩士論文庫

簡易檢索 / 詳目顯示

回結果列表

研究生：	卓訓發 Cho, Hsun-Fa
論文名稱：	以長短期記憶模型進行筆寫單字辨識 On-Pen Handwritten Word Recognition Using Long Short-Term Memory Model
指導教授：	周百祥 Chou, Pai H.
口試委員:	蔡明哲 Tsai, Ming-Jer 周志遠 Chou, Jerry
學位類別：	碩士 Master
系所名稱：
論文出版年：	2018
畢業學年度：	106
語文別：	英文
論文頁數：	32
中文關鍵詞：	嵌入式系統、筆寫辨識、手勢辨識、深度學習、連續資料分段、長短期記憶模型
外文關鍵詞：	embedded system, op-pen handwritten recognition, gesture recognition, deep learning, segmentation of continuous data, long short-term memory model
相關次數：	點閱：1 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

近年來，因穿戴式裝置的流行，基於動作感測的手寫辨識研究越來越多，例如空中手寫、智慧筆等等，但是能達到實用水平的應用卻很少。受限於感測器的限制、連續動作的分割難度這些問題，手寫動作辨識很難做有效率的文字輸入。在這篇論文中，我們提出了一種安裝在筆上的新型硬體裝置，搭配以長短期記憶模型(LSTM)為核心的辨識系統，讓使用者能夠用任何他們想拿來寫字的筆，作為一個有效率的文字輸入介面。

This thesis describes a system for text input from handwriting using a
conventional pen with a clip-on sensing unit. The clip-on unit is a
wireless sensor node that collects data from a triaxial accelerometer and a triaxial gyroscope and transmits it to a conventional personal computer. The host computer then performs segmentation to handle continuous handwriting, followed by LSTM-based classification. Moreover, we use a lexicon-based corrector to increase the accuracy. Experimental results show our proposed system to achieve good accuracy and reasonable latency for interactive use.

Contents i
Acknowledgments iv
Introduction 1
1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Related Work 3
1 Isolated Handwriting Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Continuous Handwriting Recognition . . . . . . . . . . . . . . . . . . . . . . . . . 4
System Overview 6
1 Pen Subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Host Subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Data Preprocessing and Segmentation 8
1 Gravity Elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Signal Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Data segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Our Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Recognition Method 15
1 Background: RNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Long Short-Term Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
i
3 Lexicon Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1 Bayes’ Corrector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Weighted Levenshtein distance . . . . . . . . . . . . . . . . . . . . . . . . . 19
Evaluation 21
1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.1 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2 System environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.1 Segmentation Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Accuarcy of Using Different Features . . . . . . . . . . . . . . . . . . . . . 24
2.3 Improvement After Lexicon Support . . . . . . . . . . . . . . . . . . . . . . 24
3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Conclusions and Future Work 28
1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
                                

[1] T. F. Gao and C. L. Liu, “High accuracy hand written Chinese character recognition using LDAbased
compound distances,” Pattern Recognition, vol. 41, pp. 3442–3451, 2008.
[2] R. Ebrahimzadeh and M. Jampour, “Efficient handwritten digit recognition based on histogram
of oriented gradients and SVM,” International Journal of Computer Applications, vol. 104,
2014.
[3] G. Katiyar and S. Mehfuz, “MLPNN based handwritten character recognition using combined
feature extraction,” in International Conference on Computing, Communication and Automation,
pp. 2872–2884, May 2015.
[4] C. Bahlmann, B. Haasdonk, and H. Burkhardt, “Online handwriting recognition with support
vector machines - a kernel approach,” in Proceedings of the Eighth International Workshop on
Frontiers in Handwriting Recognition (IWFHR), Aug 2002.
[5] M. Sepahvand, F. Abdali-Mohammadi, and F. Mardukhi, “Evolutionary metric-learning-based
recognition algorithm for online isolated Persian/Arabic characters, reconstructed using inertial
pen signals,” IEEE Transactions on Cybernetics, vol. 47, pp. 88–95, 2016.
[6] J. K. Oh, S.-J. Cho, W.-C. Bang, W. Chang, E. Choi, J. Yang, J. Cho, and D. Y. Kim, “Inertial
sensor based recognition of 3-D character gestures with an ensemble of classifiers,” in Proceedings
of the 9th Int’l Workshop on Frontiers in Handwriting Recognition, Oct 2004.
[7] T. Bluche, H. Ney, and C. Kermorvant, “Feature extraction with convolutional neural networks
for handwritten word recognition,” Proceedings of the 28th Benelux Conference on Artificial
Intelligence (BNAIC), pp. 285–289, 2007.
30
[8] M. Kozielski, P. Doetsch, and H. Ney, “Improvements in RWTH’s system for off-line handwriting
recognition,” in International Conference on Document Analysis and Recognition, Aug
2013.
[9] Y. Shkarupa, R. Mencis, and M. Sabatelli, “Offline handwriting recognition using LSTM recurrent
neural networks,” Proceedings of the 28th Benelux Conference on Artificial Intelligence
(BNAIC), pp. 88–95, 2016.
[10] C. Amma, M. Georgi, and T. Schultz, “Airwriting: Hands-free mobile text input by spotting and
continuous recognition of 3d-space handwriting with inertial sensors,” in International Symposium
on Wearable Computers, p. 52–59, Jun 2012.
[11] M. Chen, G. AlRegib, and B.-H. Juang, “Air-writing recognition—part I: Detection and recognition
of writing activity in continuous stream of motion data,” IEEE Transactions on Human-
Machine Systems, vol. 46, 2016.
[12] M. Chen, G. AlRegib, and B.-H. Juang, “Air-writing recognition—part II: Detection and recognition
of writing activity in continuous stream of motion data,” IEEE Transactions on Human-
Machine Systems, vol. 46, 2016.
[13] Embedded Platform Lab (EPL) at National Tsing Hua University (NTHU) in Taiwan, “Eco
Mini.” http://epl.tw/ecomini/, 2014.
[14] G. A. Abandah and F. T. Jamour, “Recognizing handwritten Arabic script through efficient
skeleton-based grapheme segmentation algorithm,” in Intelligent Systems Design and Applications,
Nov 2010.
[15] T. A. Osman, M. J. Paulik, and M. Krishnan, “An online signature verification system based on
multivariate autoregressive modeling and DTW segmentation,” in Signal Processing Applications
for Public Security and Forensics, Apr 2007.
[16] J. Sueiras, V. Ruiz, A. Sanchez, and J. F. Velez, “Offline continuous handwriting recognition
using sequence to sequence neural networks,” Neurocomputing, vol. 289, pp. 119–128, 2018.
[17] D. Saadb, B. Cohena, and E. Maroma, “Efficient training of recurrentneural network with time
delays,” Neural Networks, vol. 10, 1997.
31
[18] H. Jaeger, “A tutorial on training recurrent neural networks, covering BPPT, RTRL, EKF and
the “echo state network” approach,” in GMD Report 159, German National Research Center for
Information Technology, pp. 1–46, Oct 2002.
[19] F. A. Gers, J. Schmidhuber, and F. Cummins, “Learn to forget: Continual prediction with
LSTM,” in Technical Report IDSIA-01-99, pp. 1–20, Jan 1999.
[20] TensorFlow, “Preparing models for mobile deployment.” https://www.tensorflow.org/mobile/
prepare_models.

簡易檢索 / 詳目顯示

相關論文