One of the biggest challenges in modern societies is the improvement of healthy aging and the support to older persons in their daily activities. In particular, given its social and economic impact, the automatic detection of falls has attracted considerable attention in the computer vision and pattern recognition communities. Although the approaches based on wearable sensors have provided high detection rates, some of the potential users are reluctant to wear them and thus their use is not yet normalized. As a consequence, alternative approaches such as vision-based methods have emerged. We firmly believe that the irruption of the Smart Environments and the Internet of Things paradigms, together with the increasing number of cameras in our daily environment, forms an optimal context for vision-based systems. Consequently, here we propose a vision-based solution using Convolutional Neural Networks to decide if a sequence of frames contains a person falling. To model the video motion and make the system scenario independent, we use optical flow images as input to the networks followed by a novel three-step training phase. Furthermore, our method is evaluated in three public datasets achieving the state-of-the-art results in all three of them.

As the average age of the urban population increases, cities must adapt to improve the quality of life of their citizens. The City4Age H2020 project is working on the early detection of the risks related to mild cognitive impairment and frailty and on providing meaningful interventions that prevent these risks. As part of the risk detection process, we have developed a multilevel conceptual model that describes the user behaviour using actions, activities, and intra- and inter-activity behaviour. Using this conceptual model, we have created a deep learning architecture based on long short-term memory networks (LSTMs) that models the inter-activity behaviour. The presented architecture offers a probabilistic model that allows us to predict the user’s next actions and to identify anomalous user behaviours.

Julen Etxaniz, Gorka Azkune, Aitor Soroa, Oier Lacalle, Mikel Artetxe. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers). 2024.

Human activity recognition is a key technology for ICT-based (infomation and communication technologies) assistive applications. The most successful activity recognition systems for intelligent environments in terms of performance rely on supervised learning techniques. However, those techniques demand large labelled data sets for specific sensor deployments and monitored person. Such requirements make supervised learning techniques not to scale well to real world deployments, where different sensor infrastructures may be used to monitor different users. In this paper, we present a novel activity recognition system, based on a combination of unsupervised learning techniques and knowledge-based activity models. First, we use a domain-specific data mining algorithm previously developed by Cook et al. to extract the most frequent action sequences executed by a person. Second, we insert knowledge-based activity models in a novel matching algorithm with the aim of inferring what activities are being performed in a given action sequence. The approach results on a scalable activity recognition system, which has been tested on three real data sets. The obtained performance is comparable to supervised learning techniques.

Evaluating human activity recognition systems usually implies following expensive and time-consuming methodologies, where experiments with humans are run with the consequent ethical and legal issues. We propose a novel evaluation methodology to overcome the enumerated problems, which is based on surveys for users and a synthetic dataset generator tool. Surveys allow capturing how different users perform activities of daily living, while the synthetic dataset generator is used to create properly labelled activity datasets modelled with the information extracted from surveys. Important aspects, such as sensor noise, varying time lapses and user erratic behaviour, can also be simulated using the tool. The proposed methodology is shown to have very important advantages that allow researchers to carry out their work more efficiently. To evaluate the approach, a synthetic dataset generated following the proposed methodology is compared to a real dataset computing the similarity between sensor occurrence frequencies. It is concluded that the similarity between both datasets is more than significant.