Compensating for object variability in DNN–HMM object-centered human activity recognition

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

Standard

Compensating for object variability in DNN–HMM object-centered human activity recognition. / Peng, Yikai; Jancovic, Peter; Russell, Martin.

2019 27th European Signal Processing Conference (EUSIPCO). IEEE Xplore, 2019. (Signal Processing Conference (EUSIPCO), European).

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

Harvard

Peng, Y, Jancovic, P & Russell, M 2019, Compensating for object variability in DNN–HMM object-centered human activity recognition. in 2019 27th European Signal Processing Conference (EUSIPCO). Signal Processing Conference (EUSIPCO), European, IEEE Xplore, 27th European Signal Processing Conference, A Coruña, Spain, 2/09/19. https://doi.org/10.23919/EUSIPCO.2019.8903124

APA

Peng, Y., Jancovic, P., & Russell, M. (2019). Compensating for object variability in DNN–HMM object-centered human activity recognition. In 2019 27th European Signal Processing Conference (EUSIPCO) (Signal Processing Conference (EUSIPCO), European). IEEE Xplore. https://doi.org/10.23919/EUSIPCO.2019.8903124

Vancouver

Peng Y, Jancovic P, Russell M. Compensating for object variability in DNN–HMM object-centered human activity recognition. In 2019 27th European Signal Processing Conference (EUSIPCO). IEEE Xplore. 2019. (Signal Processing Conference (EUSIPCO), European). https://doi.org/10.23919/EUSIPCO.2019.8903124

Author

Peng, Yikai ; Jancovic, Peter ; Russell, Martin. / Compensating for object variability in DNN–HMM object-centered human activity recognition. 2019 27th European Signal Processing Conference (EUSIPCO). IEEE Xplore, 2019. (Signal Processing Conference (EUSIPCO), European).

Bibtex

@inproceedings{79c2df592fd44454a84a71c5120f6fcd,
title = "Compensating for object variability in DNN–HMM object-centered human activity recognition",
abstract = "This paper describes a deep neural network –hidden Markov model (DNN-HMM) human activity recognition system based on instrumented objects and studies compensation strategies to deal with object variability. The sensors, comprising an accelerometer, gyroscope, magnetometer and force-sensitive resistors (FSRs), are packaged in a coaster attached to the base of an object, here a mug. Results are presented for recognition of actions involved in manipulating a mug. Evaluations are performed using over 24 hours of data recordings containing sequences of actions, labelled without time-stamp information. We demonstrate the importance of data alignments. While the DNN-HMM system achieved error rate below 0.1% for matched train-test conditions, this increased up to 26.5% for highly mismatched conditions. The error rate averaged over all conditions was 1.4% when using multi-condition training and decreased to 0.8% by employing feature augmentation. The use of FSR feature compensation, specific to weight variability, resulted in 0.24% error rate. ",
keywords = "Action recognition, Compensation, DNN-HMM, Deep neural networks, Feature augmentation, Hidden Markov models, Instrumented objects, Sensors",
author = "Yikai Peng and Peter Jancovic and Martin Russell",
year = "2019",
month = nov,
day = "18",
doi = "10.23919/EUSIPCO.2019.8903124",
language = "English",
isbn = "9781538673003",
series = "Signal Processing Conference (EUSIPCO), European",
publisher = "IEEE Xplore",
booktitle = "2019 27th European Signal Processing Conference (EUSIPCO)",
note = "27th European Signal Processing Conference, EUSIPCO 2019 ; Conference date: 02-09-2019 Through 06-09-2019",

}

RIS

TY - GEN

T1 - Compensating for object variability in DNN–HMM object-centered human activity recognition

AU - Peng, Yikai

AU - Jancovic, Peter

AU - Russell, Martin

PY - 2019/11/18

Y1 - 2019/11/18

N2 - This paper describes a deep neural network –hidden Markov model (DNN-HMM) human activity recognition system based on instrumented objects and studies compensation strategies to deal with object variability. The sensors, comprising an accelerometer, gyroscope, magnetometer and force-sensitive resistors (FSRs), are packaged in a coaster attached to the base of an object, here a mug. Results are presented for recognition of actions involved in manipulating a mug. Evaluations are performed using over 24 hours of data recordings containing sequences of actions, labelled without time-stamp information. We demonstrate the importance of data alignments. While the DNN-HMM system achieved error rate below 0.1% for matched train-test conditions, this increased up to 26.5% for highly mismatched conditions. The error rate averaged over all conditions was 1.4% when using multi-condition training and decreased to 0.8% by employing feature augmentation. The use of FSR feature compensation, specific to weight variability, resulted in 0.24% error rate.

AB - This paper describes a deep neural network –hidden Markov model (DNN-HMM) human activity recognition system based on instrumented objects and studies compensation strategies to deal with object variability. The sensors, comprising an accelerometer, gyroscope, magnetometer and force-sensitive resistors (FSRs), are packaged in a coaster attached to the base of an object, here a mug. Results are presented for recognition of actions involved in manipulating a mug. Evaluations are performed using over 24 hours of data recordings containing sequences of actions, labelled without time-stamp information. We demonstrate the importance of data alignments. While the DNN-HMM system achieved error rate below 0.1% for matched train-test conditions, this increased up to 26.5% for highly mismatched conditions. The error rate averaged over all conditions was 1.4% when using multi-condition training and decreased to 0.8% by employing feature augmentation. The use of FSR feature compensation, specific to weight variability, resulted in 0.24% error rate.

KW - Action recognition

KW - Compensation

KW - DNN-HMM

KW - Deep neural networks

KW - Feature augmentation

KW - Hidden Markov models

KW - Instrumented objects

KW - Sensors

UR - http://www.scopus.com/inward/record.url?scp=85075610798&partnerID=8YFLogxK

U2 - 10.23919/EUSIPCO.2019.8903124

DO - 10.23919/EUSIPCO.2019.8903124

M3 - Conference contribution

SN - 9781538673003

T3 - Signal Processing Conference (EUSIPCO), European

BT - 2019 27th European Signal Processing Conference (EUSIPCO)

PB - IEEE Xplore

T2 - 27th European Signal Processing Conference

Y2 - 2 September 2019 through 6 September 2019

ER -