Virtual Booth

Parkinson’s Disease (PD) is a progressive neurological disorder that affects body movement, balance, posture, and walking pattern. Current gait assessment methods are commonly performed through clinical observation, which may be subjective, time-limited, and less suitable for continuous early screening. Therefore, this project aims to develop GaitAI, a wearable IMU-based gait analysis system for Parkinson’s Disease early detection using machine learning. The system uses MPU6050 IMU sensors placed on knee guards to collect tri-axial acceleration data from left and right leg movement. The ESP32 microcontroller is used to acquire sensor readings, transmit data wirelessly through the Blynk platform, and store movement data in Google Sheets for further analysis. The collected gait data are processed through data preprocessing, signal filtering, window-based feature extraction, and machine learning classification. Features such as acceleration magnitude, RMS, jerk, cadence, step frequency, dominant frequency, tremor power, gait power, and freeze ratio are extracted to represent gait behaviour. Several machine learning models were tested, including SVM, Random Forest, Decision Tree, KNN, and ANN, with the best model selected for website integration. The innovation of this project is the combination of low-cost wearable hardware, IoT-based data collection, and machine learning analysis in a single early detection platform. This system benefits users by providing a portable, affordable, and non-invasive method to support early Parkinson-like gait screening. In conclusion, GaitAI has the potential to assist healthcare workers and users in identifying abnormal gait patterns earlier, while supporting future development of smart wearable healthcare systems.

Keywords: Parkinson’s Disease, Gait Analysis, IMU Sensor, ESP32, Machine Learning