Human Activity and Motion Pattern Recognition within Indoor Environment Using Convolutional Neural Networks...

Area of Science:

• Computer Science
• Artificial Intelligence
• Signal Processing

Background:

• Human Activity Recognition (HAR) systems analyze sensor data to classify movements.
• Current HAR systems require improved responsiveness for distinguishing true from false alarms in near real-time.
• Accurate alarm triggering necessitates training systems on legitimate user motion patterns.

Purpose of the Study:

• To evaluate the accuracy and robustness of a combined Convolutional Neural Network (CNN) and Naive Bayes approach for identifying true alarm triggers.
• To assess pattern recognition capabilities using these methods, even with partial motion data.

Main Methods:

• Utilized a combined Convolutional Neural Network (CNN) and Naive Bayes model.
• Trained the system on stored motion patterns of legitimate users.
• Tested the system with diverse activity patterns, including those similar and dissimilar to training data.
• Evaluated performance based on the correct identification of true alarm triggers, such as buzzer sounds.

Main Results:

• Both the CNN and Naive Bayes approaches demonstrated effective pattern recognition for identifying true alarms.
• The methods showed robustness even when analyzing partial motion patterns derived from a full motion path.
• The combined approach improved accuracy and reliability in distinguishing legitimate user activities from false alarms.

Conclusions:

• The combined CNN and Naive Bayes model offers a viable solution for enhancing the accuracy and responsiveness of HAR systems.
• Effective pattern recognition, even with incomplete data, is achievable, leading to more reliable alarm triggering.
• This research contributes to the development of more dependable HAR systems for various applications.