To spot keywords on handwritten documents, we present a hybrid keyword spotting system, based on features extracted with Convolutional Deep Belief Networks and using Dynamic Time Warping for word scoring. Features are learned from word images, in an unsupervised manner, using a sliding window to extract horizontal patches. For two single writer historical data sets, it is shown that the proposed learned feature extractor outperforms two standard sets of features.
Titre
Keyword spotting with convolutional deep belief networks and dynamic time warping
Date
2016-09
Publié dans
Proceedings of the 25th International Conference on Artificial Neural Networks and Machine Learning (ICANN), 6-9 September 2016, Barcelona, Spain
Volume
2016, part 2
Editeur
Barcelona, Spain, 6-9 September 2016
Pagination
pp. 113-120
Présenté à
Artificial Neural Networks and Machine Learning – ICANN 2016, Barcelona, Spain, 2016-09-06, 2016-09-09
ISBN
978-3-319-44780-3
ISSN
0302-9743
Collection et n°
Lecture Notes in Computer Science (LNCS),
Type de papier
published full paper
Domaine
Ingénierie et Architecture
Ecole
HEIA-FR
Institut
iCoSys - Institut des systèmes complexes