Discovering Daily Activity Patterns from Sensor Data Sequences and Activity Sequences

Area of Science:

• Gerontology
• Computer Science
• Artificial Intelligence

Background:

• Increasing need for elderly care and independence.
• Role of technology in monitoring daily activities.
• Importance of recognizing habitual vs. uncommon behavior.

Purpose of the Study:

• Propose a novel approach to discover common daily routines of individuals.
• Develop a system to detect unusual activity sequences.
• Evaluate different methods for routine discovery and anomaly detection.

Main Methods:

• Sequence comparison and clustering for routine partitioning.
• Utilizing daily activity vectors and sensor data.
• Comparing generalized Hamming distance with Levenshtein distance.

Main Results:

• Daily activity vectors are crucial for accurate routine discovery.
• Generalized Hamming distance outperforms Levenshtein distance for sequence comparison.
• The proposed method effectively partitions daily routines for anomaly detection.

Conclusions:

• The developed approach enhances the ability to monitor elderly individuals' independence.
• Accurate routine discovery is key to detecting deviations from normal behavior.
• This technology can support proactive care and timely interventions for the elderly.