From accelerometers to circadian rhythms: applications of synchrosqueezing and wavelet coherence for assessing behavior in farm poultry

Abstract

The growing demand for welfare improvement of poultry in intensive farming systems requires strategies that go beyond monitoring productive variables, such as weight and egg-laying, and include the assessment of individual behaviour. To this end, we propose combining inertial sensors (such as accelerometers) with time-series analysis to extract biologically relevant information. Since animal behavior is naturally associated with the 24-hour daily cycle of light and darkness, in this study, we propose the use of wavelet-based approaches to evaluate the modulation of these rhythms in response to changes in the rearing environment. Specifically, we use Synchrosqueezing to characterize 24- hour rhythms in various behavioral signals derived from the acceleration vector and wavelet coherence to assess the relationship between these signals. This methodology was applied to activity and dust-bathing behaviors in Japanese quails, monitored over 11 days with acceleration data recorded at 25 Hz under three rearing substrate conditions (i.e., mesh floor, wood shavings floor, or wood shavings floor with a sandpit). Our results show that the strength of dust-bathing rhythms and the coherence between both behavioral rhythms can vary depending on the substrate. Therefore, this wavelet-based approach provides a valuable tool for the continuous monitoring of relevant behavioral rhythms in assessing animal welfare.