Predicting Human Actions Taking into Account Object Affordances

Abstract

Anticipating human intentional actions is essential for many applications involving service robots and social robots. Nowadays assisting robots must do reasoning beyond the present with predicting future actions. It is difficult due to its non-Markovian property and the rich contextual information. This task requires the subtle details inherent in human movements that may imply a future action. This paper presents a probabilistic method for action prediction in human-object interactions.The key idea of our approach is the description of the so-called object affordance, the concept which allows us to deliver a trajectory visualizing a possible future action. Extensive experiments were conducted to show the effectiveness of our method in action prediction. For evaluation we applied a new RGB-D activity video dataset recorded by the Sez3D depth sensors. The dataset contains several human activities composed out of different actions.

Prognosing Human Activity Using Actions Forecast and Structured Database

“…Furthermore, work [5] incorporated object affordances to anticipate human activities for reactive responses. Work [9] addressed the so-called, modified object affordance with spatio-temporal humanobject relation taking into account selected features. This research showed that careful selection of features and proper data gathering approaches are crucial for forecast performances, cancelling the need for a sophisticated learning algorithm.…”

“…A limited number of studies [4], [5], [9] addresses relevant factors such as spatial features, human-object interactions, data gathering, and the database. Although activity recognition and forecasting methods have shown good performance for small datasets with controlled background settings, it is still challenging to generalize them for real-time and uncontrolled settings due to large computational complexity.…”

“…Although activity recognition and forecasting methods have shown good performance for small datasets with controlled background settings, it is still challenging to generalize them for real-time and uncontrolled settings due to large computational complexity. Our previous work [9] presented some situations with higher computational need, however, still on a limited scale. Visual information may be redundant, and therefore it is not necessary to consider each observation in the data processing.…”

Complex Human–Object Interactions Analyzer Using a DCNN and SVM Hybrid Approach

“…The authors of paper [4] proposed the anticipation of human intentional actions using the affordance of objects and the context of scenes for visualizing possible future actions. The experiments were performed using a Sez3D sensor.…”

“…Next, we calculated the Euclidean distance between two consecutive frames for all joints using Equation (2). Then we found the mean and the standard deviation for joint distance as described in Equations (3) and (4).…”