Background: Parkinson's disease (PD) is a neurodegenerative and progress disease. The late detection of PD causes low quality of life of patients. Therefore, the system to diagnose the disease in the early stage, which can be used outside the hospitals, is necessary. Previous studies have shown that PD patients and controls can be differentiated in several gait parameters and classified using machine learning. However, high accurate prediction of early PD and healthy has not been achieved.

Background: Normal pressure hydrocephalus (NPH) is characterized by three symptoms, disturbance of gait, dementia and urinary incontinence. Especially, disturbance of gait is known as a relatively early symptom. Differential diagnosis of NPH is not easy since many symptoms overlap with other neurological diseases. The cerebrospinal fluid shunting is expected to ameliorate symptoms of NPH. The late detection of disease may limit the treatment effect. Therefore, the early detection of NPH is important. Previous studies have reported that clinical features of the gait abnormality in NPH. These measurements need relatively large instruments, such as motion capture systems.