The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Uma estrutura de fusão entre CNN e RNN é proposta especificamente para reconhecimento de escrita aérea. Ao modelar a escrita aérea usando recursos espaciais e temporais, a rede proposta pode aprender mais informações do que as técnicas existentes. O desempenho da rede proposta é avaliado usando conjuntos de dados alfabéticos e numéricos no banco de dados público, nomeadamente o 6DMG. A precisão média da rede de fusão proposta supera outras técnicas, ou seja, 99.25% e 99.83% são observados no gesto alfabético e no gesto numérico, respectivamente. Também é proposta uma estrutura simplificada de RNN, que pode atingir cerca de duas vezes a velocidade da rede BLSTM comum. Confirma-se também que apenas a distância entre pontos de amostragem consecutivos é suficiente para atingir um alto desempenho de reconhecimento.
Buntueng YANA
Osaka University
Takao ONOYE
Osaka University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Buntueng YANA, Takao ONOYE, "Air-Writing Recognition Based on Fusion Network for Learning Spatial and Temporal Features" in IEICE TRANSACTIONS on Fundamentals,
vol. E101-A, no. 11, pp. 1737-1744, November 2018, doi: 10.1587/transfun.E101.A.1737.
Abstract: A fusion framework between CNN and RNN is proposed dedicatedly for air-writing recognition. By modeling the air-writing using both spatial and temporal features, the proposed network can learn more information than existing techniques. Performance of the proposed network is evaluated by using the alphabet and numeric datasets in the public database namely the 6DMG. Average accuracy of the proposed fusion network outperforms other techniques, i.e. 99.25% and 99.83% are observed in the alphabet gesture and the numeric gesture, respectively. Simplified structure of RNN is also proposed, which can attain about two folds speed-up of ordinary BLSTM network. It is also confirmed that only the distance between consecutive sampling points is enough to attain high recognition performance.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.E101.A.1737/_p
Copiar
@ARTICLE{e101-a_11_1737,
author={Buntueng YANA, Takao ONOYE, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Air-Writing Recognition Based on Fusion Network for Learning Spatial and Temporal Features},
year={2018},
volume={E101-A},
number={11},
pages={1737-1744},
abstract={A fusion framework between CNN and RNN is proposed dedicatedly for air-writing recognition. By modeling the air-writing using both spatial and temporal features, the proposed network can learn more information than existing techniques. Performance of the proposed network is evaluated by using the alphabet and numeric datasets in the public database namely the 6DMG. Average accuracy of the proposed fusion network outperforms other techniques, i.e. 99.25% and 99.83% are observed in the alphabet gesture and the numeric gesture, respectively. Simplified structure of RNN is also proposed, which can attain about two folds speed-up of ordinary BLSTM network. It is also confirmed that only the distance between consecutive sampling points is enough to attain high recognition performance.},
keywords={},
doi={10.1587/transfun.E101.A.1737},
ISSN={1745-1337},
month={November},}
Copiar
TY - JOUR
TI - Air-Writing Recognition Based on Fusion Network for Learning Spatial and Temporal Features
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1737
EP - 1744
AU - Buntueng YANA
AU - Takao ONOYE
PY - 2018
DO - 10.1587/transfun.E101.A.1737
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E101-A
IS - 11
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - November 2018
AB - A fusion framework between CNN and RNN is proposed dedicatedly for air-writing recognition. By modeling the air-writing using both spatial and temporal features, the proposed network can learn more information than existing techniques. Performance of the proposed network is evaluated by using the alphabet and numeric datasets in the public database namely the 6DMG. Average accuracy of the proposed fusion network outperforms other techniques, i.e. 99.25% and 99.83% are observed in the alphabet gesture and the numeric gesture, respectively. Simplified structure of RNN is also proposed, which can attain about two folds speed-up of ordinary BLSTM network. It is also confirmed that only the distance between consecutive sampling points is enough to attain high recognition performance.
ER -