The study of skeleton description reduction in the human fall-detection task

Abstract

Accurate and reliable real-time fall detection is a key aspect of any intelligent elderly people care system. A lot of modern RGB-D cameras can provide a skeleton description of a human figure as a compact pose presentation. This makes it possible to use this description for further analysis without access to real video and, thus, to increase the privacy of the whole system. The skeleton description reduction based on the anthropometrical characteristics of a human body is proposed. The experimental study on the TST Fall Detection dataset v2 by the Leave-One-Person-Out method shows that the proposed skeleton description reduction technique provides better recognition quality and increases the overall performance of a Fall-Detection System.

The basic assembly of skeletal models in the fall detection problem

“…Next, we splitted the dataset into training and test parts, with the condition that the training set includes ten of the eleven actors in the database, and the remaining actor record was used exclusively for tests. Such a test procedure was applied to each actor (Leave-one-person-out) [9,10].…”

“…Nowadays, along with the growing interest in research of human behavior and activity, the problem of representing a human figure with a skeleton obtained from RGB-D sensors is also becoming relevant. Accordingly, there are a large number of approaches to the construction of 2D and 3D skeletal models [9,10,35,36]. In previous study of the fall detection problem [10] a skeletal model (Fig.…”

“…A comprehensive overview of databases could be found in [29]. In the works [9,10] dedicated to the human fall detection problem, the TEST Fall Detection v2 database [30] was used for experimental research of fall detection algorithms. It contains depth maps and 3D skeletons collected by the Microsoft Kinect v2 sensor and presented as records of various durations.…”

Near real-time animal action recognition and classification

“…Одна из актуальных задач компьютерного зрения -задача распознавания действий объекта на основании его изображения или серии изображений [1 -3]. Проводить распознавание действий при решении такой задачи можно как у механических [4], так и у живых объектов (люди [5,6], животные [7], насекомые [3]). Решение рассматриваемой задачи необходимо для обеспечения возможности автоматизированного наблюдения за объектом и принятия необходимых решений на основании полученной информации [8].…”