Intelligent Diagnosis of Motor Nervous System Diseases & Multi-sensor-based Human Motion Recognition
Photo by OpenPose
Advisor: Yanan Sui; Associate Professor in School of Aerospace Engineering, Tsinghua University; National Engineering Laboratory for Neuromodulation.
Project Area: Machine learning, Computer Vision.
- We established a more stable and reliable automatic grading system for Parkinson’s Disease (PD) patients’ online diagnosis and grading through machine learning classification based on action videos of PD patients.
- Collected keypoints by using the open-source OpenPose human body gesture recognition project, extracted hundreds of features on both time and frequency domain after filtering, and then used SVM (Support Vector Machine) to achieve automatic grading of patients' disease. Mastered multiple methods of data-processing, feature extraction and models of supervised learning.
- The multi-sensor-based human motion recognition part of the project the extension part that is being conducted. I am attempting to establish a set of algorithms to automatically improve the serious core issue of missing and mis-recognizing trace points which greatly limited the application of motion capture.
