An Enhanced Human Activity Recognition Algorithm with Positional Attention

Chenyang Xu, Jianfei Shen, Feiyi Fan, Tian Qiu, Zhihong Mao
Proceedings of The 14th Asian Conference on Machine Learning, PMLR 189:1181-1196, 2023.

Abstract

Human activity recognition (HAR) attracts widespread attention from researchers recently, and deep learning is employed as a dominant paradigm of solving HAR problems. The previous techniques rely on domain knowledge or attention mechanism extract long-range dependency in temporal dimension and cross channel correlation in sensor’s channel dimension. In this paper, a HAR model with positional attention (PA), termed as PA-HAR, is presented. To enhance the features in both sensor’s channel and temporal dimensions, we propose to split the sensor signals into two 1D features to capture the long-range dependency along the temporal-axis and signal’s cross-channel information along the sensor’s channel-axis. Furthermore, we embed the features with positional information by encoding the generated features into pairs of temporal-aware and sensor’s channel-aware attention maps and weighting the input feature maps. Extensive experiments based on five public datasets demonstrate that the proposed PA-HAR algorithm achieves a competitive performance in HAR related tasks compared with the state-of-the-art approaches.

Cite this Paper

BibTeX
@InProceedings{pmlr-v189-xu23a, title = {An Enhanced Human Activity Recognition Algorithm with Positional Attention}, author = {Xu, Chenyang and Shen, Jianfei and Fan, Feiyi and Qiu, Tian and Mao, Zhihong}, booktitle = {Proceedings of The 14th Asian Conference on Machine Learning}, pages = {1181--1196}, year = {2023}, editor = {Khan, Emtiyaz and Gonen, Mehmet}, volume = {189}, series = {Proceedings of Machine Learning Research}, month = {12--14 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v189/xu23a/xu23a.pdf}, url = {https://proceedings.mlr.press/v189/xu23a.html}, abstract = {Human activity recognition (HAR) attracts widespread attention from researchers recently, and deep learning is employed as a dominant paradigm of solving HAR problems. The previous techniques rely on domain knowledge or attention mechanism extract long-range dependency in temporal dimension and cross channel correlation in sensor’s channel dimension. In this paper, a HAR model with positional attention (PA), termed as PA-HAR, is presented. To enhance the features in both sensor’s channel and temporal dimensions, we propose to split the sensor signals into two 1D features to capture the long-range dependency along the temporal-axis and signal’s cross-channel information along the sensor’s channel-axis. Furthermore, we embed the features with positional information by encoding the generated features into pairs of temporal-aware and sensor’s channel-aware attention maps and weighting the input feature maps. Extensive experiments based on five public datasets demonstrate that the proposed PA-HAR algorithm achieves a competitive performance in HAR related tasks compared with the state-of-the-art approaches.} }
Endnote
%0 Conference Paper %T An Enhanced Human Activity Recognition Algorithm with Positional Attention %A Chenyang Xu %A Jianfei Shen %A Feiyi Fan %A Tian Qiu %A Zhihong Mao %B Proceedings of The 14th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Emtiyaz Khan %E Mehmet Gonen %F pmlr-v189-xu23a %I PMLR %P 1181--1196 %U https://proceedings.mlr.press/v189/xu23a.html %V 189 %X Human activity recognition (HAR) attracts widespread attention from researchers recently, and deep learning is employed as a dominant paradigm of solving HAR problems. The previous techniques rely on domain knowledge or attention mechanism extract long-range dependency in temporal dimension and cross channel correlation in sensor’s channel dimension. In this paper, a HAR model with positional attention (PA), termed as PA-HAR, is presented. To enhance the features in both sensor’s channel and temporal dimensions, we propose to split the sensor signals into two 1D features to capture the long-range dependency along the temporal-axis and signal’s cross-channel information along the sensor’s channel-axis. Furthermore, we embed the features with positional information by encoding the generated features into pairs of temporal-aware and sensor’s channel-aware attention maps and weighting the input feature maps. Extensive experiments based on five public datasets demonstrate that the proposed PA-HAR algorithm achieves a competitive performance in HAR related tasks compared with the state-of-the-art approaches.
APA
Xu, C., Shen, J., Fan, F., Qiu, T. & Mao, Z.. (2023). An Enhanced Human Activity Recognition Algorithm with Positional Attention. Proceedings of The 14th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 189:1181-1196 Available from https://proceedings.mlr.press/v189/xu23a.html.

Related Material