Design and Development of an Imitation Detection System for Human Action Recognition Using Deep Learning

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Human Action Recognition (HAR) is crucial for applications in surveillance, sports analysis, and healthcare.
• Accurate classification of human actions from video data is a key challenge in HAR.

Purpose of the Study:

• To design and develop an imitation detection system using deep learning-based HAR algorithms.
• To explore the efficacy of deep learning models for accurate human action classification.

Main Methods:

• Implemented and evaluated deep learning models: a single-frame Convolutional Neural Network (CNN) and a pretrained VGG-16 model.
• Compared the performance of deep learning models against classical classifiers: K-Nearest Neighbors, Support Vector Machine, and Random Forest.
• Utilized the benchmark KTH dataset for model evaluation.

Main Results:

• The VGG-16 model demonstrated superior performance, achieving a 98% accuracy rate.
• VGG-16 significantly outperformed the single-frame CNN model in action classification.
• Deep learning approaches showed higher accuracy compared to classical classifiers.

Conclusions:

• Deep learning models, particularly VGG-16, are highly effective for human action recognition and imitation detection.
• The developed system offers a promising solution for applications requiring accurate human action classification.
• Further research can explore advanced deep learning architectures for enhanced HAR performance.