Advanced internet of things enhanced activity recognition for disability people using deep learning model with...

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Human activity recognition (HAR) is crucial for applications like virtual reality, surveillance, and home monitoring.
• Existing HAR models are being developed for disabled individuals, leveraging data from devices like smartphones and cameras.
• HAR utilizes AI, deep learning, and machine learning to identify and classify human actions.

Purpose of the Study:

• To propose an Enhanced Activity Recognition for Disability People Using a Deep Learning Model and Nature-Inspired Optimization Algorithms (EARDP-DLMNOA) model.
• To improve HAR model performance through advanced optimization algorithms.
• To enhance activity recognition specifically for disabled individuals.

Main Methods:

• Data normalization using min-max normalization.
• Feature subset selection via adaptive chimp optimization (AdCO).
• Activity recognition using a deep convolutional auto-encoder (DCAE) with hyperparameter optimization by the zebra optimization algorithm (ZOA).

Main Results:

• The EARDP-DLMNOA model demonstrated superior performance in human activity recognition.
• Achieved a high accuracy of 97.58% on the HAR through the smartphone dataset.
• Outperformed existing methods in experimental validation.

Conclusions:

• The EARDP-DLMNOA model effectively enhances activity recognition for disabled individuals.
• The integration of deep learning and nature-inspired optimization algorithms significantly improves HAR accuracy.
• This approach offers a promising solution for advanced human activity recognition in assistive technologies.