TY - GEN
T1 - A static hand gesture recognition for peruvian sign language using digital image processing and deep learning
AU - Lazo, Cristian
AU - Sanchez, Zaid
AU - del Carpio, Christian
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - The work consists in recognizing the gestures of the alphabet in Peruvian sign language using techniques of digital image processing and a model of Deep Learning (CNN). Image processing techniques are used for segmentation and tracking of the hand of the person making the gestures. Once the image of the segmented hand is used, a CNN classification model is used to be able to recognize the gesture. The image processing and CNN algorithms were implemented in the Python programming language. The database used was 23,000 images divided into 70% for training, 15% for testing and 15% for validation. Likewise, said data corresponds to 1000 images for each non-mobile gesture of the alphabet. The results obtained for the precision of the classifier were 99.89, 99.88 and 99.85% for the data of training, test and validation respectively. In the case of the Log Loss parameter, 0.0132, 0.0036, and 0.0107 were obtained for the training, testing and validation data, respectively.
AB - The work consists in recognizing the gestures of the alphabet in Peruvian sign language using techniques of digital image processing and a model of Deep Learning (CNN). Image processing techniques are used for segmentation and tracking of the hand of the person making the gestures. Once the image of the segmented hand is used, a CNN classification model is used to be able to recognize the gesture. The image processing and CNN algorithms were implemented in the Python programming language. The database used was 23,000 images divided into 70% for training, 15% for testing and 15% for validation. Likewise, said data corresponds to 1000 images for each non-mobile gesture of the alphabet. The results obtained for the precision of the classifier were 99.89, 99.88 and 99.85% for the data of training, test and validation respectively. In the case of the Log Loss parameter, 0.0132, 0.0036, and 0.0107 were obtained for the training, testing and validation data, respectively.
KW - Deep learning
KW - Digital image processing
KW - Python
KW - Static hand gesture
UR - https://www.scopus.com/pages/publications/85068599499
U2 - 10.1007/978-3-030-16053-1_27
DO - 10.1007/978-3-030-16053-1_27
M3 - Contribución a la conferencia
AN - SCOPUS:85068599499
SN - 9783030160524
T3 - Smart Innovation, Systems and Technologies
SP - 281
EP - 290
BT - Proceedings of the 4th Brazilian Technology Symposium (BTSym’18) - Emerging Trends and Challenges in Technology
A2 - Iano, Yuzo
A2 - Loschi, Hermes José
A2 - Arthur, Rangel
A2 - Saotome, Osamu
A2 - Vieira Estrela, Vânia
PB - Springer Science and Business Media Deutschland GmbH
T2 - 4th Brazilian Technology Symposium, BTSym 2018
Y2 - 23 October 2018 through 25 October 2018
ER -