在這篇論文中我們提出了一個採用投票機制的卷積神經網路，應用在手
寫文件的語言辨識上，為了可以取得文件中更細部的資訊，我們會對文件
的每一行進行切割，再對每一行進行滑動窗口的分割，之後對每個滑動窗
口去進行卷積神經網路的特徵提取，然後用SOFTMAX 層來進行分類。在
ICDAR2011 公開的數據集的實驗中，我們的方法取得了很高的辨識率，在
比較複雜的公開數據集Multilingual Handwritten Dataset 上，也取得一個不錯
的成果，比起傳統對一整張文件可以提取出更多有用的特徵資訊，效能上也
有明顯比較高的辨識率。

In this paper, we propose a language identification method use voting base Convolutional
Neural Networks approach. We apply line extraction and sliding windows
before CNN feature extraction. After CNN feature extraction, CNN feature
classify by softmax layer. In experiment, we get high performance on ICDAR2011
public dataset, on more complex dataset, Multilingual Handwritten Dataset also has
a good performance. Compared to full-page feature extraction, our method would
take more useful information.

[1] University of maryland, laboratory for language and media processing (lamp)
, handwritten language and writer id dataset. http://lamp.cfar.umd.edu,
2016.
[2] J. Almazán, A. Gordo, A. Fornés, and E. Valveny. Word spotting and recognition
with embedded attributes. Pattern Analysis and Machine Intelligence,
IEEE Transactions on, 36(12):2552–2566, 2014.
[3] N. Arvanitopoulos and S. Süsstrunk. Seam carving for text line extraction
on color and grayscale historical manuscripts. In Frontiers in Handwriting
Recognition (ICFHR), 2014 14th International Conference on, pages 726–731.
IEEE, 2014.
[4] S. Avidan and A. Shamir. Seam carving for content-aware image resizing. In
ACM Transactions on graphics (TOG), volume 26, page 10. ACM, 2007.
[5] M. Bresler. Text/non-text classification of strokes using the composite descriptor.
In Proceedings of the 17th International Student Conference on Electrical
Engineering (POSTER2013), Prague, Czech, pages 1–5, 2013.
[6] L. Chen, S. Wang, W. Fan, J. Sun, and N. Satoshi. Deep learning based language
and orientation recognition in document analysis. In Document Analysis
and Recognition (ICDAR), 2015 13th International Conference on, pages
436–440. IEEE, 2015.
[7] E. Indermuhle, V. Frinken, and H. Bunke. Mode detection in online handwritten
documents using blstm neural networks. In Frontiers in Handwriting
Recognition (ICFHR), 2012 International Conference on, pages 302–307.
IEEE, 2012.
[8] M. Liwicki, E. Indermuhle, and H. Bunke. On-line handwritten text line detection
using dynamic programming. In Ninth International Conference onDocument Analysis and Recognition (ICDAR 2007), volume 1, pages 447–
451. IEEE, 2007.
[9] G. Louloudis, N. Stamatopoulos, and B. Gatos. Icdar 2011 writer identification
contest. In 2011 International Conference on Document Analysis and
Recognition, pages 1475–1479. IEEE, 2011.
[10] A. Nicolaou, A. Bagdanov, L. Gomez-Bigorda, and D. Karatzas. Visual script
and language identification. arXiv preprint arXiv:1601.01885, 2016.
[11] T. Ojala, M. Pietikäinen, and T. Mäenpää. Multiresolution gray-scale and rotation
invariant texture classification with local binary patterns. Pattern Analysis
and Machine Intelligence, IEEE Transactions on, 24(7):971–987, 2002.
[12] R. Sarkar, S. Moulik, N. Das, S. Basu, M. Nasipuri, and M. Kundu. Suppression
of non-text components in handwritten document images. In Image
Information Processing (ICIIP), 2011 International Conference on, pages 1–7.
IEEE, 2011.
[13] B. Shi, X. Bai, and C. Yao. Script identification in the wild via discriminative
convolutional neural network. Pattern Recognition, 52:448–458, 2016.
[14] A. Ul-Hasan, M. Z. Afzal, F. Shafait, M. Liwicki, and T. M. Breuel. A sequence
learning approach for multiple script identification. In Document Analysis
and Recognition (ICDAR), 2015 13th International Conference on, pages
1046–1050. IEEE, 2015.