VOLUME 13 NUMBER 2 (July to December 2020)

Anton Louise P. De Ocampo*¹ and Elmer P. Dadios²

¹Electronics and Communications Engineering Department,
      De La Salle University, Manila, Philippines
²Manufacturing Engineering and Management Department
      De La Salle University, Manila, Philippines

Human activity recognition (HAR) systems can be categorized as a sensor- or vision-based depending on the type of data it collects as inputs. The non-contact implementation of vision-based HAR is the confounding factor why such systems are preferred over systems using wearable sensors. Although remarkable feats have been achieved in the field of human activity recognition, one challenge remains the focus of recent researches –extraction and selection of suitable features. In this work, a novel approach in extracting and selecting feature vector for HAR implementation in smart farming is proposed. By exploiting the data distribution on stacks of difference maps, a new feature vector, which contains high discriminative attributes between activities performed by farmers in the field, is proposed. Using the k-NN classifier, the experiment obtained 98.89%, 98.69%, and 98.79% scores in precision, recall, and F1-measure in the classification of farmers’ activities in the field.