An Effective Approach for Human Activity Classification Using Feature Fusion and Machine Learning Methods.

Area of Science:

• Computer Vision
• Machine Learning
• Image Processing

Background:

• Human activity classification is crucial for applications like medical informatics, surveillance, and human-computer interaction.
• Existing methods using machine learning (ML) and soft computational algorithms have limitations.
• Advanced computer vision techniques are needed for robust human activity classification from video sequences.

Purpose of the Study:

• To propose an effective automated approach for human activity classification from video frames.
• To enhance the accuracy and performance of human activity recognition systems.

Main Methods:

• The proposed approach involves five steps: preprocessing, feature extraction, feature selection, feature fusion, and classification.
• Machine learning (ML) classifiers were trained, validated, and tested on two public benchmark datasets.
• Feature fusion and ML methods were employed for automated classification.

Main Results:

• The developed approach achieved high accuracies of 99.5% and 99.9% on two benchmark datasets.
• The method demonstrated superior performance compared to existing approaches.
• High evaluation metrics, including sensitivity, accuracy, precision, and specificity, were attained.

Conclusions:

• The proposed automated approach using feature fusion and ML is effective for human activity classification.
• This method offers a promising direction for advancing computer vision-based activity recognition.
• The high performance indicates potential for real-world applications in healthcare and surveillance.