Blind Modalities for Human Activity Recognition.

Area of Science:

• Computer Science
• Artificial Intelligence
• Signal Processing

Background:

• Human Activity Recognition (HAR) is crucial for automation in smart homes, assisted living, and security.
• Optical modalities (RGB, depth, thermal) dominate HAR, leveraging deep learning (CNNs) and large datasets.
• Visual privacy concerns and limitations in constrained environments drive the need for alternative HAR methods.

Purpose of the Study:

• To provide a comprehensive overview of the Human Activity Recognition (HAR) modalities landscape.
• To discuss emerging unconventional sensing modalities for HAR.
• To inform researchers and practitioners about challenges and opportunities in HAR.

Main Methods:

• Literature review of Human Activity Recognition (HAR) modalities.
• Analysis of emerging unconventional sensing technologies (radar, WiFi, seismic, environmental sensors).
• Discussion of novel applications and unconventional uses of these modalities.

Main Results:

• Optical modalities are currently prevalent but have limitations.
• Emerging modalities offer contactless, long-range sensing with enhanced privacy.
• Unconventional applications demonstrate the potential of these new HAR approaches.

Conclusions:

• Emerging unconventional modalities present significant opportunities for advancing Human Activity Recognition (HAR).
• Further research is needed to address challenges and fully exploit the potential of non-optical HAR sensors.
• The field of HAR is expanding beyond traditional imaging techniques.