Recognition of Activities of Daily Living with Egocentric Vision: A Review

Abstract

Video-based recognition of activities of daily living (ADLs) is being used in ambient assisted living systems in order to support the independent living of older people. However, current systems based on cameras located in the environment present a number of problems, such as occlusions and a limited field of view. Recently, wearable cameras have begun to be exploited. This paper presents a review of the state of the art of egocentric vision systems for the recognition of ADLs following a hierarchical structure: motion, action and activity levels, where each level provides higher semantic information and involves a longer time frame. The current egocentric vision literature suggests that ADLs recognition is mainly driven by the objects present in the scene, especially those associated with specific tasks. However, although object-based approaches have proven popular, object recognition remains a challenge due to the intra-class variations found in unconstrained scenarios. As a consequence, the performance of current systems is far from satisfactory.

Comparison of fiducial marker detection and object interaction in activities of daily living utilising a wearable vision sensor

“…Occlusions that may be created through environmental objects, such as doors and large items of furniture, or occlusions generated by the user themselves, such as hands/head/torso occluding objects that they are interacting with [34]. One of the methods biggest strengths is the ease to which it is able to be deployed within differing environments, the use of fiducial markers with associated ID's negates the need for specific training to each environment.…”

“…One of the methods biggest strengths is the ease to which it is able to be deployed within differing environments, the use of fiducial markers with associated ID's negates the need for specific training to each environment. As the object is the centre of the users attention, the object is usually in the centre of the image and in focus, providing a high quality image for processing [34]. Secondly, the use of a moving camera couple with static objects reduces the issues traditionally seen with a static camera solution such as the limited field of view, which may require the installation of multiple cameras within an environment.…”

“…Secondly, the use of a moving camera couple with static objects reduces the issues traditionally seen with a static camera solution such as the limited field of view, which may require the installation of multiple cameras within an environment. Occlusions that may be created through environmental objects, such as doors and large items of furniture, or occlusions generated by the user themselves, such as hands/head/torso occluding objects that they are interacting with [34]. This coupled with being a superior solution for object interaction due to the added advantages a head-mounted camera provides.…”

Jointly Learning Energy Expenditures and Activities Using Egocentric Multimodal Signals

“…Recent surveys can be found in [48,8,5,30], which address various tasks such as activity recognition, object recognition, and video summarization. Here, we focus on activity recognition using egocentric video and wearable sensors.…”

“…More recently, egocentric first-person cameras have gained popularity as a new modality of wearable sensors. Activity recognition in visual data is a widely studied problem in computer vision [22,63,35,2], and a number of works have investigated this task in the domain of egocentric video [48,8,5,30] and in combination with other wearable sensors [59,44].…”