Employment of Ensemble Machine Learning Methods for Human Activity Recognition

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Human activity and gesture recognition are crucial for real-time applications.
• While sensors like accelerometers and gyroscopes aid motion detection, software-based methods are needed for unique gesture classification.
• Machine learning algorithms can distinguish human actions but face challenges with large datasets and model complexity.

Purpose of the Study:

• To develop an efficient and accurate human activity and gesture recognition system.
• To address the complexity and efficiency issues in machine learning models for human action classification.
• To leverage ensemble learning methods for improved performance in distinguishing human gestures.

Main Methods:

• Implemented ensemble learning by combining multiple trained machine learning models.
• Preprocessed the dataset using Principal Component Analysis (PCA) for dimensionality reduction.
• Applied Synthetic Minority Oversampling Technique (SMOTE) and K-means clustering to optimize the dataset and reduce complexity while preserving feature importance.

Main Results:

• Achieved a maximum accuracy of 99.36% using ensemble methods.
• Demonstrated significant reduction in dataset complexity through preprocessing techniques.
• Validated the effectiveness of stacking and voting ensemble approaches for human gesture classification.

Conclusions:

• Ensemble learning methods, combined with effective data preprocessing, significantly enhance the accuracy and efficiency of human activity and gesture recognition.
• The proposed approach successfully overcomes the challenges posed by large datasets and model complexity.
• The research offers a robust solution for real-time human motion and action prediction.