CNN based approach for activity recognition using a wrist-worn accelerometer.

Area of Science:

• Human-Computer Interaction
• Machine Learning
• Wearable Technology

Background:

• Conventional human activity recognition relies heavily on manual feature engineering, a complex and computationally intensive process.
• Optimal feature selection in traditional methods is challenging and time-consuming.
• Deep learning offers a promising alternative by automating feature extraction.

Purpose of the Study:

• To develop a generalized deep learning model for recognizing fundamental human forearm movements.
• To reduce computational costs associated with feature engineering in activity recognition.
• To validate the model's performance under various data conditions.

Main Methods:

• A novel deep learning framework was employed for automatic feature extraction.
• Data was collected from four subjects using a single wrist-worn accelerometer sensor.
• The model was validated using different pre-processing techniques and noisy data conditions.

Main Results:

• The proposed deep learning methodology achieved an average recognition rate of 99.8%.
• This performance significantly outperformed conventional methods like K-means clustering, linear discriminant analysis, and support vector machine.
• The model demonstrated robustness across different data pre-processing and noise levels.

Conclusions:

• Deep learning effectively automates feature extraction for human activity recognition.
• The proposed model offers a highly accurate and computationally efficient solution for forearm movement recognition.
• This approach represents a significant advancement over traditional machine learning techniques in this domain.