Activityp atternso fd airy cattle have receivedi ncreasingi nteresti nr ecenty earsb ecause they promisei nsightsi ntoh ealth statea nd well-being.T he fusion with data from additionals ensors ignals promises ac omprehensive monitoringofactivitypatternscomposed of sequences of single activitystates. We used a combinationo faSupportV ector Machine( SVM),astateo ft he artc lassification method, andaConditionalR andomF ield (CRF).S VMsd istinguish single states, whereasC RFsl abel states equences

more »

... nderc onsideration of specified constraints. In ap reliminary experiment,aLocal Positioning System wasc ombinedw ith a heartr ates ensori no rder to estimates even spatiotemporal activity states.T he applicationo ft he CRF to theS VM result causedaslight increase in accuracy (5%) but am ajor improvementa tt he correct determinationo fl ong sequences (increasingl engtho ft he longestc ommons ubsequencef rom3 481t o6 207 periods). This robust detectiono fl ong lyings equences allowedf or theu naffected extractiono f therestingpulse.