The image perspective correction methods for Parkinson's scale actions assessment

Parkinson's Disease is a neurodegenerative disorder commonly classified as a movement disorder. The symptoms of Parkinson's can be assessed through various types of testing methods. Compared to Parkinson's assessment methods based on various human body sensors, computer vision-based assessment methods are more convenient and efficient. In a home environment, however, the camera cannot always maintain a standard posture, which often leads to poor image quality. Therefore, this paper proposes an image correction method based on human body bounding boxes. The corrective effect of the method on the patient's scale movements from various perspectives was tested by comparing it with benchmark data from the Vicon system. The average pixel error has been reduced from over 80 to below 10, and the correction effectiveness rate is now over 80%. This method enables the assessment of Parkinson's symptoms to be conducted at home using only a smartphone. It offers robust support for non-invasive, real-time assessment of Parkinson's symptoms, providing potential convenience for daily care and remote monitoring of Parkinson's patients.