Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM)

Research output: Chapter in Book/Report/Conference proceedingConference contribution

11 Citations (Scopus)

Abstract

Recurrent neural networks (RNN) are powerful sequence learners. However, RNN suffers from the problem of vanishing gradient point. This fact makes learning sequential task more than 10 time steps harder for RNN. Recurrent network with LSTM cells as hidden layers (LSTM-RNN) is a deep learning recurrent network architecture designed to address the vanishing gradient problem by incorporating memory cells (LSTM cells) in the hidden layer(s). This advantage puts it at one of the best sequence learners for time-series data such as cursive hand writings, protein structure prediction, speech recognition and many more task that require learning through long time lags [2][3][4], In this paper, we applied the concept of using recurrent networks with LSTM cells as hidden layer to learn the behaviours of a humanoid robot based on multiple sequences of joint data from 10 joints on the NAO robot. We show that the LSTM network is able to learn the patterns in the data and effectively classify the sequences into 6 different trained behaviors.

Original languageEnglish
Title of host publication2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages109-114
Number of pages6
ISBN (Electronic)9781479957651
DOIs
Publication statusPublished - 09 Oct 2015
EventIEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014 - Kuala Lumpur, Malaysia
Duration: 15 Dec 201416 Dec 2014

Publication series

Name2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014

Other

OtherIEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014
CountryMalaysia
CityKuala Lumpur
Period15/12/1416/12/14

Fingerprint

Recurrent neural networks
Robots
Network architecture
Speech recognition
Time series
Long short-term memory
Proteins
Data storage equipment

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Industrial and Manufacturing Engineering
  • Mechanical Engineering

Cite this

How, D. N. T., Sahari, K. S. M., Yuhuang, H., & Kiong, L. C. (2015). Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM). In 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014 (pp. 109-114). [7295871] (2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ROMA.2014.7295871
How, Dickson Neoh Tze ; Sahari, Khairul Salleh Mohamed ; Yuhuang, Hu ; Kiong, Loo Chu. / Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM). 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 109-114 (2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014).
@inproceedings{750ea2dfed6941168fd9e9b684875f48,
title = "Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM)",
abstract = "Recurrent neural networks (RNN) are powerful sequence learners. However, RNN suffers from the problem of vanishing gradient point. This fact makes learning sequential task more than 10 time steps harder for RNN. Recurrent network with LSTM cells as hidden layers (LSTM-RNN) is a deep learning recurrent network architecture designed to address the vanishing gradient problem by incorporating memory cells (LSTM cells) in the hidden layer(s). This advantage puts it at one of the best sequence learners for time-series data such as cursive hand writings, protein structure prediction, speech recognition and many more task that require learning through long time lags [2][3][4], In this paper, we applied the concept of using recurrent networks with LSTM cells as hidden layer to learn the behaviours of a humanoid robot based on multiple sequences of joint data from 10 joints on the NAO robot. We show that the LSTM network is able to learn the patterns in the data and effectively classify the sequences into 6 different trained behaviors.",
author = "How, {Dickson Neoh Tze} and Sahari, {Khairul Salleh Mohamed} and Hu Yuhuang and Kiong, {Loo Chu}",
year = "2015",
month = "10",
day = "9",
doi = "10.1109/ROMA.2014.7295871",
language = "English",
series = "2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "109--114",
booktitle = "2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014",
address = "United States",

}

How, DNT, Sahari, KSM, Yuhuang, H & Kiong, LC 2015, Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM). in 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014., 7295871, 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014, Institute of Electrical and Electronics Engineers Inc., pp. 109-114, IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014, Kuala Lumpur, Malaysia, 15/12/14. https://doi.org/10.1109/ROMA.2014.7295871

Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM). / How, Dickson Neoh Tze; Sahari, Khairul Salleh Mohamed; Yuhuang, Hu; Kiong, Loo Chu.

2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014. Institute of Electrical and Electronics Engineers Inc., 2015. p. 109-114 7295871 (2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM)

AU - How, Dickson Neoh Tze

AU - Sahari, Khairul Salleh Mohamed

AU - Yuhuang, Hu

AU - Kiong, Loo Chu

PY - 2015/10/9

Y1 - 2015/10/9

N2 - Recurrent neural networks (RNN) are powerful sequence learners. However, RNN suffers from the problem of vanishing gradient point. This fact makes learning sequential task more than 10 time steps harder for RNN. Recurrent network with LSTM cells as hidden layers (LSTM-RNN) is a deep learning recurrent network architecture designed to address the vanishing gradient problem by incorporating memory cells (LSTM cells) in the hidden layer(s). This advantage puts it at one of the best sequence learners for time-series data such as cursive hand writings, protein structure prediction, speech recognition and many more task that require learning through long time lags [2][3][4], In this paper, we applied the concept of using recurrent networks with LSTM cells as hidden layer to learn the behaviours of a humanoid robot based on multiple sequences of joint data from 10 joints on the NAO robot. We show that the LSTM network is able to learn the patterns in the data and effectively classify the sequences into 6 different trained behaviors.

AB - Recurrent neural networks (RNN) are powerful sequence learners. However, RNN suffers from the problem of vanishing gradient point. This fact makes learning sequential task more than 10 time steps harder for RNN. Recurrent network with LSTM cells as hidden layers (LSTM-RNN) is a deep learning recurrent network architecture designed to address the vanishing gradient problem by incorporating memory cells (LSTM cells) in the hidden layer(s). This advantage puts it at one of the best sequence learners for time-series data such as cursive hand writings, protein structure prediction, speech recognition and many more task that require learning through long time lags [2][3][4], In this paper, we applied the concept of using recurrent networks with LSTM cells as hidden layer to learn the behaviours of a humanoid robot based on multiple sequences of joint data from 10 joints on the NAO robot. We show that the LSTM network is able to learn the patterns in the data and effectively classify the sequences into 6 different trained behaviors.

UR - http://www.scopus.com/inward/record.url?scp=84959472628&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84959472628&partnerID=8YFLogxK

U2 - 10.1109/ROMA.2014.7295871

DO - 10.1109/ROMA.2014.7295871

M3 - Conference contribution

AN - SCOPUS:84959472628

T3 - 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014

SP - 109

EP - 114

BT - 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

How DNT, Sahari KSM, Yuhuang H, Kiong LC. Multiple sequence behavior recognition on humanoid robot using long short-term memory (LSTM). In 2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014. Institute of Electrical and Electronics Engineers Inc. 2015. p. 109-114. 7295871. (2014 IEEE International Symposium on Robotics and Manufacturing Automation, IEEE-ROMA2014). https://doi.org/10.1109/ROMA.2014.7295871