Related Experiment Video
Updated: Aug 10, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
3.9K
Sensor-Based Activity Recognition Using Frequency Band Enhancement Filters and Model Ensembles.
Hyuga Tsutsumi1, Kei Kondo1, Koki Takenaka1
1Graduate School of Engineering, University of Fukui, Fukui 910-8507, Japan.
Sensors (Basel, Switzerland)
|February 11, 2023
Summary
This study enhances sensor-based activity recognition using frequency filters and ensemble learning. The proposed method significantly improves accuracy by emphasizing important frequency bands in accelerometer data.
Area of Science:
- Human-Computer Interaction
- Signal Processing
- Machine Learning
Background:
- Sensor-based activity recognition commonly uses deep learning with accelerometer and gyroscope data.
- While frequency spectrum analysis is used, data augmentation based on frequency characteristics remains underexplored.
- Existing methods may not fully leverage the frequency domain for improved activity recognition accuracy.
Purpose of the Study:
- To propose a novel activity recognition method combining ensemble learning with frequency enhancement filters.
- To identify and emphasize critical frequency bands in accelerometer data for specific activities.
- To evaluate the effectiveness and robustness of the proposed method against existing approaches.
Main Methods:
- Experimentally identified important frequency bands by masking accelerometer data and assessing accuracy changes.
- Developed an activity recognition model incorporating ensemble learning and frequency enhancement filters.
- Validated the method using four diverse datasets, comparing performance with and without enhancement filters and ensemble learning.
Main Results:
- The proposed method, utilizing frequency band enhancement filters during training and testing alongside ensemble learning, achieved the highest recognition accuracy.
- Comparison across four datasets demonstrated the robustness of the approach, with the proposed method outperforming single models in three out of four cases.
- The integration of frequency-specific filters and ensemble techniques proved more effective than traditional methods.
Conclusions:
- Frequency band enhancement filters and ensemble learning significantly boost accuracy in sensor-based activity recognition.
- The proposed method offers a robust and effective approach for human activity recognition using accelerometer data.
- Further investigation into frequency characteristics can unlock new potential for advanced sensor-based recognition systems.
Related Concept Videos
01:03IR Frequency Region: Fingerprint Region
1.0K
IR spectra are divided into two main regions: the diagnostic region and the fingerprint region. The diagnostic region of the spectrum lies above 1500 cm−1. The absorptions resulting from single-bond vibrations of the N–H, C–H, and O–H stretch at higher wavenumbers and appear on the left side of the spectrum. The stretching absorptions of the C≡C and C≡N occur between 2100–2300 cm−1. In contrast, those arising from stretching absorptions of the...
1.0K
01:30Classification of Signals
620
In signal processing, signals are classified based on various characteristics: continuous-time versus discrete-time, periodic versus aperiodic, analog versus digital, and causal versus noncausal. Each category highlights distinct properties crucial for understanding and manipulating signals.
A continuous-time signal holds a value at every instant in time, representing information seamlessly. In contrast, a discrete-time signal holds values only at specific moments, often denoted as x(n), where...
A continuous-time signal holds a value at every instant in time, representing information seamlessly. In contrast, a discrete-time signal holds values only at specific moments, often denoted as x(n), where...
620