| Proj No. | A3253-251 |
| Title | LLM-Powered Human Activity Recognition (HAR) Using Wearable IMU Sensors |
| Summary | Objective The project focuses on enhancing human activity recognition (HAR) using IMU (inertial measurement unit) sensor data, enabling LLMs to classify and reason about human movements in real-world applications such as healthcare and fitness. Scope Use public HAR datasets (e.g., UCI Smartphone-Based Activity Recognition) with accelerometer and gyroscope data. Implement IoT-LLM techniques to preprocess IMU signals into LLM-friendly formats. Utilize chain-of-thought prompting and RAG to improve activity classification accuracy and reasoning. Develop an LLM-powered analytics dashboard that explains why certain movements were detected and suggests fitness or rehabilitation recommendations. Required Skills Python, NumPy, and data processing for time-series IMU data. Knowledge of basic deep learning models for activity recognition. Familiarity with OpenAI API / LangChain for reasoning and report generation. |
| Supervisor | Ast/P Yang Jianfei (Loc:School of Mechanical and Aerospace Engineering (MAE), Ext: ?) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Internet of Things Laboratory (Loc: S1-B4c-14, ext: 5470/5475) |
| Single/Group: | Single |
| Area: | Digital Media Processing and Computer Engineering |
| ISP/RI/SMP/SCP?: |