Transition Activity Recognition System based on Standard Deviation Trend Analysis.

Area of Science:

• Computer Science
• Engineering
• Signal Processing

Background:

• Sensor-based human activity recognition (HAR) is increasingly prevalent due to advancements in micro-electromechanical systems (MEMS) and smartphones.
• Existing HAR systems often struggle with accurately identifying transition activities, the brief periods between distinct actions.

Purpose of the Study:

• To develop and evaluate a novel algorithm for the accurate recognition of transition activities in HAR.
• To address the limitations of current HAR methods that do not specifically handle transitional phases.

Main Methods:

• The study proposes a new algorithm based on standard deviation trend analysis (STD-TA).
• This method specifically targets the identification of transition activities, differentiating them from basic activities.

Main Results:

• The proposed STD-TA algorithm demonstrates superior performance in recognizing transition activities.
• The system achieved an overall accuracy exceeding 80% when tested on real-world data.

Conclusions:

• The STD-TA algorithm offers a significant improvement for HAR systems by accurately identifying transition activities.
• This approach enhances the overall performance and robustness of sensor-based human activity recognition.