Multimodal Sequential Modeling and Recognition of Human Activities
Video-based recognition of activities of daily living (ADLs) is being used in ambient assisted living systems in order to support independent living of old people. In this work, we propose a new multimodal ADL recognition method by modeling the correlatio
- PDF / 806,527 Bytes
- 8 Pages / 439.37 x 666.14 pts Page_size
- 79 Downloads / 174 Views
)
UR: SAGE - Systèmes Avancés en Génie Electrique, Université de Sousse, Sousse, Tunisia [email protected] 2 SAMOVAR, Telecom SudParis, CNRS, University Paris Saclay, Palaiseau, France [email protected]
Abstract. Video-based recognition of activities of daily living (ADLs) is being used in ambient assisted living systems in order to support independent living of old people. In this work, we propose a new multimodal ADL recognition method by modeling the correlation between motion and object information. We encode motion using dense interest point trajectories which are robust to occlusion and speed variability. We formulate the learning problem using a two-layer SVM hidden conditional random field (HCRF) recognition model that is particularly relevant for multimodal sequence recognition. This hierarchical classifier opti‐ mally combines the discriminative power of SVM and the long-range feature dependencies modeling by the HCRF. Keywords: Ambient assisted living system · Activities of daily living · Multimodal representation · Interest points · SVM-HCRF
1
Introduction
According to the World Health Organization, the number of elderly people is going to reach 2 billion by 2050. On the other hand, 89 % of seniors prefer to stay in their own homes. Such a demographic and social context brings significant challenges for health care systems and society in terms of increased costs of nursing home care and lack of resources. As an alternative to current care models, ambient assisted living (AAL) systems can offer the possibility for many old and vulnerable people to live independ‐ ently and safely at home. One of the most important components in AAL systems is the activity of daily living (ADL) recognition component. ADL recognition helps evaluating the degree of dependency of the elderly by detecting changes in their behavior patterns, identifying early signs of dementia [1] and detecting their critical situations such as falls, which can enable early elderly assistance. ADLs that affect the independence of elders include taking medications, eating, bathing, dressing, cleaning, and socializing. A wide-range of ADL recognition systems in the context of AAL applications use sensors either embedded in the environment or body worn. Recently, a growing popu‐ larity has been noted towards video sensors based monitoring of ADLs in AAL systems [2, 3] as they provide richer sensory information than traditional sensors. These sensors have relatively a low cost compared to systems based on environment sensors which © Springer International Publishing Switzerland 2016 K. Miesenberger et al. (Eds.): ICCHP 2016, Part II, LNCS 9759, pp. 541–548, 2016. DOI: 10.1007/978-3-319-41267-2_76
542
M. Selmi and M.A. El-Yacoubi
require generally a large network of sensors, which makes them relatively obtrusive and costly to maintain. However, video based AALs systems have several obstacles including users’ acceptance and the risk of loss of privacy. In spite of advances on vision techniques, recognizing ADLs in natural
Data Loading...
