Shallow hierarchical CNN–LSTM for activity recognition to integrate postural transition states

Doug Tilley; Uriel Martinez Hernandez

doi:10.1109/SENSORS56945.2023.10325044

Shallow hierarchical CNN–LSTM for activity recognition to integrate postural transition states

Doug Tilley, Uriel Martinez Hernandez

Research output: Contribution to conference › Paper › peer-review

4 Citations (SciVal)

Abstract

This paper presents an approach to enhance classification accuracy of Human Activity Recognition (HAR) datasets using a shallow hierarchical method, comprised of 2 Convolutional layers and 1 LSTM Layer, to identify transition tasks. The designed model was tested on the Human Activity and Postural Transition (HAPT) dataset from the University of California, Irvine, achieving an accuracy of 93 %, demonstrating it's efficacy in correctly identifying a variety of human activities, including the postural transition states. An F1 score of 85 % was attained, highlighting a reliable method of reducing false positives on an unbalanced dataset. The performance on these metrics showcases the model's reliability and proficiency in classifying activities within unbalanced datasets.

Original language	English
Number of pages	4
DOIs	https://doi.org/10.1109/SENSORS56945.2023.10325044
Publication status	Published - 23 Nov 2023
Event	IEEE Sensors 2023 - Hilton Vienna Park, Vienna, Austria Duration: 29 Oct 2023 → 1 Nov 2023 https://doi.org/10.1109/SENSORS56945.2023

Conference

Conference	IEEE Sensors 2023
Country/Territory	Austria
City	Vienna
Period	29/10/23 → 1/11/23
Internet address	https://doi.org/10.1109/SENSORS56945.2023

Funding

This work is supported by the UKRI Centre for Doctoral Training in Accountable, Responsible and Transparent AI (ART-AI), under UKRI grant number EP/S023437/1.

Funders	Funder number
UK Research & Innovation	EP/S023437/1

Keywords

Human Activity Recognition
CNN-LSTM
Hierarchical Networks
HAPT
Postural Transition

Access to Document

10.1109/SENSORS56945.2023.10325044Licence: Unspecified

Cite this

@conference{2ad288cf14384ef89f09093e96584e6d,

title = "Shallow hierarchical CNN–LSTM for activity recognition to integrate postural transition states",

abstract = "This paper presents an approach to enhance classification accuracy of Human Activity Recognition (HAR) datasets using a shallow hierarchical method, comprised of 2 Convolutional layers and 1 LSTM Layer, to identify transition tasks. The designed model was tested on the Human Activity and Postural Transition (HAPT) dataset from the University of California, Irvine, achieving an accuracy of 93 \%, demonstrating it's efficacy in correctly identifying a variety of human activities, including the postural transition states. An F1 score of 85 \% was attained, highlighting a reliable method of reducing false positives on an unbalanced dataset. The performance on these metrics showcases the model's reliability and proficiency in classifying activities within unbalanced datasets.",

keywords = "Human Activity Recognition, CNN-LSTM, Hierarchical Networks, HAPT, Postural Transition",

author = "Doug Tilley and \{Martinez Hernandez\}, Uriel",

year = "2023",

month = nov,

day = "23",

doi = "10.1109/SENSORS56945.2023.10325044",

language = "English",

note = "IEEE Sensors 2023 ; Conference date: 29-10-2023 Through 01-11-2023",

url = "https://doi.org/10.1109/SENSORS56945.2023",

}

TY - CONF

T1 - Shallow hierarchical CNN–LSTM for activity recognition to integrate postural transition states

AU - Tilley, Doug

AU - Martinez Hernandez, Uriel

PY - 2023/11/23

Y1 - 2023/11/23

N2 - This paper presents an approach to enhance classification accuracy of Human Activity Recognition (HAR) datasets using a shallow hierarchical method, comprised of 2 Convolutional layers and 1 LSTM Layer, to identify transition tasks. The designed model was tested on the Human Activity and Postural Transition (HAPT) dataset from the University of California, Irvine, achieving an accuracy of 93 %, demonstrating it's efficacy in correctly identifying a variety of human activities, including the postural transition states. An F1 score of 85 % was attained, highlighting a reliable method of reducing false positives on an unbalanced dataset. The performance on these metrics showcases the model's reliability and proficiency in classifying activities within unbalanced datasets.

AB - This paper presents an approach to enhance classification accuracy of Human Activity Recognition (HAR) datasets using a shallow hierarchical method, comprised of 2 Convolutional layers and 1 LSTM Layer, to identify transition tasks. The designed model was tested on the Human Activity and Postural Transition (HAPT) dataset from the University of California, Irvine, achieving an accuracy of 93 %, demonstrating it's efficacy in correctly identifying a variety of human activities, including the postural transition states. An F1 score of 85 % was attained, highlighting a reliable method of reducing false positives on an unbalanced dataset. The performance on these metrics showcases the model's reliability and proficiency in classifying activities within unbalanced datasets.

KW - Human Activity Recognition

KW - CNN-LSTM

KW - Hierarchical Networks

KW - HAPT

KW - Postural Transition

UR - https://www.scopus.com/pages/publications/85179758634

U2 - 10.1109/SENSORS56945.2023.10325044

DO - 10.1109/SENSORS56945.2023.10325044

M3 - Paper

T2 - IEEE Sensors 2023

Y2 - 29 October 2023 through 1 November 2023

ER -

Shallow hierarchical CNN–LSTM for activity recognition to integrate postural transition states

Abstract

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this