This dataset contains preprocessed accelerometer and gyroscope data collected from ankle-mounted sensors worn on the right ankle of people with Parkinson's Disease. The measurements were conducted in a semi-free living environment, enabling the analysis of real-world movement patterns. The data is structured into tensors for machine learning research on freezing of gait (FOG) detection. Two distinct datasets are provided: All-Activities and FOG: Includes data from various activities along with FOG episodes.Walking-Turning and FOG: Includes data specifically from walking, turning, and FOG episodes.Each dataset includes time-series data and corresponding binary labels indicating the presence or absence of FOG. Associated metadata provides additional descriptive information for each sample.