Sufficient sleep is important for calf welfare, but so far the only possibilities measuring sleep in production systems are either using ambulatory EEG or validated sleeping behaviour. We aimed at developing a wireless actigraph for measuring sleep in calves. We attached lightweight, wireless 3-dimensional accelerometers to collars of six under 6 week-old calves, kept in a straw-bedded group pen. The device registered movement and inclination in three axes at 25 Hz with sensitivity of 2g. We filmed calves behaviour simultaneously for 48 hours, and estimated the daily duration for sleep (S), rapid-eye-movement sleep (REMS) and non-REM sleep (NREMS) based on resting posture (Hänninen et al. Appl. Anim. Behav. Sci. 111, 2008). A model for predicting sleep based on the accelerometer data was developed with the data from 3 calves, and the accuracy of the model was tested with the other 3 calves. The model used the mean, variance and the wavelet variance of the data to predict sleep with a Support Vector Machine classifier. The daily time spent in S, NREMS and REMS was calculated from the observed and predicted behaviours, and compared with a t-test.
The model was able to distinguish (mean ± S.E) 93 ± 3% of total S time, 89 ± 9% of total NREMS and 83 ± 2% of total REMS. There were no statistically significant
differences (p>0.05) in daily S, REMS or NREMS between the observed and predicted behaviour for the 3 calves used in validation (621 vs. 602 min, 351 vs. 339 min, 316 vs. 351 min, respectively). We were able to record calves´ sleep using a wireless accelerometer together with a model for classifying the behavior with a good accuracy. The device provides a method to continuously measure sleep in calves in production environment without disturbing the animals.