PoseMotive is a wearable posture-monitoring system designed to help users improve their posture and body language through continuous feedback and data analysis. The system consists of 5 inertial measurement unit (IMU) sensors embedded in a wearable garment that measure body orientation and motion. A microcontroller gathering all the data from these sensors and aggregating the information and forwards it to a web application via bluetooth. The application communicates to a backend system through wireless communication which processes the sensor data and classifies the user's posture into interpretable categories such as slouched, straight, leaning left, leaning right, and arm positioning such as open or closed. The overall goal of the project is to promote better ergonomic habits by helping users become aware of posture issues and correct them through consistent monitoring and feedback.
PoseMotive is Web application complete with real time data fed into it by a custom made vest.
The frontend application was built using React. The backend infrastructure was built using FastAPI a Python-based web framework designed for high- performance APIs.
The hardware component of PoseMotive consists of a wearable garment designed to hold five BMI160 IMU sensors positioned at key locations on the body, including the left arm, left shoulder, upper back, right arm, and right shoulder. These sensor placements allow the system to capture the orientation and motion of critical body segments that influence posture. By fixing sensors in consistent anatomical locations, the system can more accurately detect posture changes and body alignment patterns. Each IMU continuously collects motion and orientation data, which is transmitted to a central hub for aggregation and synchronization. The central hub uses a Seeed Studio XIAO nRF52840 microcontroller that collects sensor readings from the IMUs, synchronizes the data, and packages it for transmission to the backend system. The microcontroller supports Bluetooth Low Energy communication, allowing the wearable sensors to transmit data wirelessly. The collected data is then sent to the backend system where machine learning models classify posture into labels such as slouched, straight, leaning left, leaning right, and arm positioning categories.
Tests focused on confirming that the sensors could detect posture variations such as slouching, leaning left or right, and changes in arm position. During these tests, users wore the prototype garment and intentionally adopted different posture positions while the sensors collected motion data. The backend system successfully received and processed this data, allowing the posture classification pipeline to interpret the orientation information and assign posture labels.
Another experiment evaluated the performance of the web application and backend infrastructure. This involved testing the authentication system, device management features, and session tracking functionality. The login system was tested to ensure that authentication tokens were generated correctly and that protected routes prevented unauthorized access to system features. Tests also confirmed that users could create sessions and view posture monitoring pages without encountering errors in the application workflow.
Testing of posture classification models also demonstrated promising results. Using collected sensor data, the system was able to distinguish between several posture categories including slouched posture, upright posture, and lateral leaning positions. Classification results show that the system can successfully detect posture changes when sensors are placed consistently on the body. These results demonstrate that the PoseMotive architecture can support posture monitoring and analysis, although additional training data and testing will be required to fully deploy the AI model in production environments.