Precision farming is based on the continuous recording of parameters on animals or their environment in order to detect deviations from the norm and correct them quickly. Until now, the objective of precision farming has essentially been to improve the efficiency of livestock farms, in particular by monitoring production parameters and animal consumption. Animal welfare, a major societal concern, is very little taken into account. We have developed a method based on the Fourier transform to detect changes in the activity rate of cows (the activity being estimated from a positioning sensor). These changes appear when the animal is stressed or sick (or even precede the appearance of clinical signs) and reflect the animal's feelings. Before putting this method into production in tools marketed by Precision Livestock Farming companies, it should be improved to limit the number of false positives (20% at present) and to try to link the form of the rhythm modification to its origin (stress, infectious disease, lameness, heat, etc.). We propose to do this by analyzing the detected modifications more finely using the Fourier transform, by exploring the wavelet method and by applying a fuzzy logic approach (in particular to take into account the progressive aspect of the modifications linked to the development and remission of a disease). Ultimately, the results should allow better consideration of the state of animal welfare in farms, beneficial not only for the animal but also for production.