Activity Recognition Using an Egocentric Perspective of Everyday Objects
نویسندگان
چکیده
Recognizing activities based on an actor’s object manipulation is an important research approach within ubiquitous computing. We present an approach which complements object manipulation with an actor’s situational information by viewing the everyday objects used by the actor to perform his/her activities from an “egocentric perspective”. Two concepts, namely observable space and manipulable space, are introduced as part of a situative space model inspired by the situated action theory to capture the changes in the set of objects seen and in the set of objects touchable by an actor in recognizing activities. A detailed evaluation of our prototype activity recognition system in virtual-reality environment is presented as a “proof of concept”. We obtained a recognition precision of 89% on the activity-level and 76% on the action-level among 10 everyday home activities using our situative space model. Virtual reality was used as a test-bed in order to speed up the design process, compensate for the limitations with currently available sensing technologies and to compare the contributions of observable space, manipulable space and object manipulation.
منابع مشابه
Egocentric Pose Recognition in Four Lines of Code
We tackle the problem of estimating the 3D pose of an individual’s upper limbs (arms+hands) from a chest mounted depth-camera. Importantly, we consider pose estimation during everyday interactions with objects. Past work shows that strong pose+viewpoint priors and depth-based features are crucial for robust performance. In egocentric views, hands and arms are observable within a well defined vo...
متن کاملObject-Centric Spatio-Temporal Pyramids for Egocentric Activity Recognition
Activities in egocentric video are largely defined by the objects with which the camera wearer interacts, making representations that summarize the objects in view quite informative. Beyond simply recording how frequently each object occurs in a single histogram, spatio-temporal binning approaches can capture the objects’ relative layout and ordering. However, existing methods use hand-crafted ...
متن کاملAn Activity-Centered Wearable Computing Infrastructure for Intelligent Environment Applications
There are many research efforts that focus on converting everyday environments into intelligent and computationally active environments that support and enhance the abilities of its occupants in executing their activities. Such environments must have the ability to recognize the activities performed by its occupant, maintain a real-time model of the environment, address the occupant’s privacy a...
متن کاملTrajectory aligned features for first person action recognition
Egocentric videos are characterised by their ability to have the first person view. With the popularity of Google Glass and GoPro, use of egocentric videos is on the rise. Recognizing action of the wearer from egocentric videos is an important problem. Unstructured movement of the camera due to natural head motion of the wearer causes sharp changes in the visual field of the egocentric camera c...
متن کاملActivity Recognition Based on Intra and Extra Manipulation of Everyday Objects
Recognizing activities based on an actor’s interaction with everyday objects is an important research approach within ubiquitous computing. We present a recognition approach which complement objects grabbed or released information with the object’s internal state changes (as an effect of intra manipulation) and the object’s external state changes with reference to other objects (as an effect of...
متن کامل