Multicondition training for PD diagnosis with voice features from diadochokinesis test

The dataset includes acoustic features extracted from voice recordings using a diadochokinesis (DDK) test protocol. Realistic environmental noise was introduced into the recordings prior to feature extraction to simulate real-world conditions. The dataset comprises voice samples from 30 Parkinson's disease (PD) patients and 30 healthy controls. All samples were captured via smartphone without audio compression to preserve data quality.