Predicting sleep and lying time of calves with a support vector machine classifier using accelerometer data

Sleep is essential to calves, but to date the only possibilities for measuring sleep in cattle production systems use ambulatory EEG or validated sleeping behavior assessments. We developed a small, neck-based, wireless accelerometer system for measuring the sleep and lying time of calves. We collected data from 10 dairy calves and developed a model based on wavelet analysis with a support vector machine classifier for measuring sleep and lying time and were able to record sleep and lying time accurately. For total sleeping time the model was able to distinguish (mean ± SE) 90 ± 3% and 85 ± 4% of the sleeping bouts, and 82 ± 2% of the occurrence of sleep. Correspondingly, the model distinguished 66 ± 8% and 70 ± 6% of the total time for NREM and REM sleep. 70 ± 6% of the NREM sleep bout lengths and 80 ± 5% of the REM sleep bouts were predicted. The numbers for NREM and REM bouts were 77 ± 5% and 79 ± 4%, respectively. The model correctly predicted 96 ± 1% of total lying time, 79 ± 6% of lying bout durations, and 77 ± 7% of the occurrence of lying bouts. The device provides a method to measure sleep and lying time in calves continuously in a production environment without disturbing the animals.