Intelligent Localization and Deep Human Activity Recognition through IoT Devices

Area of Science:

• Ubiquitous Computing
• Internet of Things (IoT)
• Human-Computer Interaction

Background:

• Ubiquitous computing is a growing research area with applications in healthcare, behavior analysis, and personal safety.
• Human activity recognition and localization are key applications within ubiquitous computing, often utilizing sensor data from wearable devices.
• Existing methods require robust models capable of processing complex sensor data for accurate recognition and localization.

Purpose of the Study:

• To propose a robust model for simultaneous human activity recognition and localization using IoT data.
• To leverage smartphone and smartwatch sensor data for enhanced user monitoring and context awareness.
• To improve the accuracy and reliability of activity recognition and localization systems.

Main Methods:

• A signal denoising technique using a second-order Butterworth filter and a Hamming window for data chunking.
• Generation of multiple stacked windows for reliable feature extraction in parallel human activity recognition and human localization modules.
• Application of recursive feature elimination and a genetic algorithm for feature selection and data augmentation, followed by training a deep neural decision forest.

Main Results:

• The proposed model achieved 88.25% accuracy for activity recognition and 90.63% for localization on the ExtraSensory dataset.
• On the Sussex-Huawei Locomotion dataset, the system demonstrated 96.00% accuracy for activity recognition and 90.50% for localization.
• The system outperformed existing state-of-the-art methods in both activity recognition and localization tasks.

Conclusions:

• The developed model provides a robust and accurate solution for recognizing human activities and classifying locations using wearable sensor data.
• The parallel processing of activity recognition and localization, combined with advanced feature engineering and deep learning, significantly enhances system performance.
• This research contributes to the field of ubiquitous computing by offering a high-performing system applicable to various real-world scenarios.