Activity identification using body-mounted sensors--a review of classification techniques

Area of Science:

• Biomedical Engineering
• Human Movement Analysis
• Wearable Technology

Background:

• Miniaturized body-worn sensing technology allows for continuous data collection on human movement in free-living conditions.
• This enables the development of automated activity profiling systems capable of long-term monitoring.
• Such systems rely on effective classification algorithms to interpret sensor data.

Purpose of the Study:

• To review various techniques used for classifying normal human activities and identifying falls from body-worn sensor data.
• To provide an overview of different analytical approaches applied in this field.
• To identify areas for future research, particularly in advanced classification methods.

Main Methods:

• Review of existing literature on classification techniques for body-worn sensor data.
• Categorization of methods based on analytical techniques.
• Illustration of diverse approaches applied in activity recognition and fall detection.

Main Results:

• Significant progress has been made in classifying activities and detecting falls using body-worn sensors.
• A variety of analytical techniques have been employed, showcasing diverse approaches.
• The field benefits from the continuous record of activity patterns over extended periods.

Conclusions:

• Body-worn sensor technology is crucial for automated activity profiling and fall detection.
• Current classification algorithms show promise but require further advancement.
• Future work should focus on applying advanced classification techniques to complex scenarios with numerous activities.