Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Activity and Subject Detection for UCI HAR Dataset with & without missing Sensor Data

Debashish Saha, Piyush Malik, Adrika Saha

TL;DR

This work presents a lightweight LSTM-based framework for simultaneous activity and subject recognition on the UCI HAR dataset, explicitly addressing missing sensor data via simple and KNN imputations, with and without PCA. The model achieves 93.89% accuracy in activity recognition and 80.19% in subject recognition, establishing a new baseline for the latter. Missing data simulations reveal substantial performance degradation without imputation, which is largely recovered by imputation schemes, particularly KNNImputer, though PCA sometimes reduces gains. The combination of LSTM, imputation, and dimensionality reduction demonstrates robustness for real-world HAR tasks where sensor outages are common, and the authors provide open-source code to foster further development.

Abstract

Current studies in Human Activity Recognition (HAR) primarily focus on the classification of activities through sensor data, while there is not much emphasis placed on recognizing the individuals performing these activities. This type of classification is very important for developing personalized and context-sensitive applications. Additionally, the issue of missing sensor data, which often occurs in practical situations due to hardware malfunctions, has not been explored yet. This paper seeks to fill these voids by introducing a lightweight LSTM-based model that can be used to classify both activities and subjects. The proposed model was used to classify the HAR dataset by UCI [1], achieving an accuracy of 93.89% in activity recognition (across six activities), nearing the 96.67% benchmark, and an accuracy of 80.19% in subject recognition (involving 30 subjects), thereby establishing a new baseline for this area of research. We then simulate the absence of sensor data to mirror real-world scenarios and incorporate imputation techniques, both with and without Principal Component Analysis (PCA), to restore incomplete datasets. We found that K-Nearest Neighbors (KNN) imputation performs the best for filling the missing sensor data without PCA because the use of PCA resulted in slightly lower accuracy. These results demonstrate how well the framework handles missing sensor data, which is a major step forward in using the Human Activity Recognition dataset for reliable classification tasks.

Activity and Subject Detection for UCI HAR Dataset with & without missing Sensor Data

TL;DR

This work presents a lightweight LSTM-based framework for simultaneous activity and subject recognition on the UCI HAR dataset, explicitly addressing missing sensor data via simple and KNN imputations, with and without PCA. The model achieves 93.89% accuracy in activity recognition and 80.19% in subject recognition, establishing a new baseline for the latter. Missing data simulations reveal substantial performance degradation without imputation, which is largely recovered by imputation schemes, particularly KNNImputer, though PCA sometimes reduces gains. The combination of LSTM, imputation, and dimensionality reduction demonstrates robustness for real-world HAR tasks where sensor outages are common, and the authors provide open-source code to foster further development.

Abstract

Current studies in Human Activity Recognition (HAR) primarily focus on the classification of activities through sensor data, while there is not much emphasis placed on recognizing the individuals performing these activities. This type of classification is very important for developing personalized and context-sensitive applications. Additionally, the issue of missing sensor data, which often occurs in practical situations due to hardware malfunctions, has not been explored yet. This paper seeks to fill these voids by introducing a lightweight LSTM-based model that can be used to classify both activities and subjects. The proposed model was used to classify the HAR dataset by UCI [1], achieving an accuracy of 93.89% in activity recognition (across six activities), nearing the 96.67% benchmark, and an accuracy of 80.19% in subject recognition (involving 30 subjects), thereby establishing a new baseline for this area of research. We then simulate the absence of sensor data to mirror real-world scenarios and incorporate imputation techniques, both with and without Principal Component Analysis (PCA), to restore incomplete datasets. We found that K-Nearest Neighbors (KNN) imputation performs the best for filling the missing sensor data without PCA because the use of PCA resulted in slightly lower accuracy. These results demonstrate how well the framework handles missing sensor data, which is a major step forward in using the Human Activity Recognition dataset for reliable classification tasks.
Paper Structure (11 sections, 12 equations, 4 figures, 2 tables)

This paper contains 11 sections, 12 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. LSTM Framework
  5. Handling Missing Sensor Data
  6. Dimensionality Reduction with PCA
  7. Dataset and Experiments
  8. Evaluation
  9. Insights from Missing Data Experiments
  10. Results
  11. Conclusion

Figures (4)

  • Figure 1: LSTM Network Architecture
  • Figure 2: Subject recognition accuracy per activity
  • Figure 3: Accuracy for activity recognition with Imputations. The blue bars represent baseline results with missing data, and the red dotted line indicates results without missing data.
  • Figure 4: Accuracy for subject recognition with Imputations. The blue bars represent baseline results with missing data, and the red dotted line indicates results without missing data.