Package of motion sensor data of spin alert behaviour

Seizure alert dogs can offer early warnings of seizures to individuals with epilepsy, yet existing approaches to using alert dogs rely on spontaneous behaviours that are difficult to validate or replicate. Our study presents a wearable seizure alert detection collar designed to recognise signalling behaviours in trained assistance dogs using machine learning and motion sensors. Data were collected from six trained dogs performing a standardised spin alert behaviour, producing 135 labelled spin events. The dataset was curated over a three-month period, resulting in the collection of 412.6 hours of motion sensor data and 3,078 seconds of synchronized video footage. The motion data was captured using an inertial measurement unit (IMU) on a smart collar, sampling at 50 Hz across six channels: three axes of linear acceleration (Accel_LN_X​, Accel_LN_Y​, Accel_LN_Z​) and three axes of angular velocity (Gyro_X​, Gyro_Y​, Gyro_Z​).To establish ground truth for model training, a meticulous manual annotation process was undertaken by cross-referencing sensor data streams with the corresponding video recordings. The final annotated dataset comprised over 3,000 seconds of time-series data, encompassing 135 distinct instances of the target spinning behaviour. These spin events accounted for a total of 349 seconds, with individual event durations ranging from 1.02 to 5.42 seconds (s). The data were collected from six canine subjects, exhibiting a notable imbalance in the distribution of events per subject: Stuart (n=74), Rosie (n=23), Teddy (n=16), Ranger (n=10), Tori (n=9), and Nadia (n=3).