Przedstawiono ideę systemu wizyjnego umożliwiającego analizę obrazów z wielu kamer dla potrzeb systemów rzeczywistości wspomaganej. Przedstawione rozwiązanie zapewnia efektywne wyznaczanie pozycji i ścieżek ruchu poruszających się osób w przestrzeni miejskiej oraz ich aktywności. W artykule przedstawiono kluczowe moduły analizy obrazu z wielu widoków, w tym: podsystem wykrywania i śledzenia pieszych (MOT) oraz podsystem wykrywania aktywności użytkowników. Omawiane podsystemy zostały przebadane w celu określenia efektywności zaproponowanych algorytmów określania ścieżki ruchu i rozpoznawania aktywności.
EN
This article presents the idea of a vision system that allows for the analysis of images from multiple cameras for the purposes of augmented reality systems. The proposed solution effectively determines the position and paths of movement of moving people in the urban space and their activities. The article presents the critical modules of the image analysis from multiple views, including the pedestrian detection and tracking (MOT) subsystem and the user activity detection subsystem. The subsystems were stud to determine the effectiveness of the proposed algorithms for determining the traffic path and recognizing activity.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The recognition of human activities is a topic of great relevance due to its wide range of applications. Different approaches have been proposed to recognize human activities, ranging from the comparison of signals with thresholds to the application of deep and machine learning techniques. In this work, the classification of six human activities (walking, walking downstairs, walking upstairs, standing, sitting, and lying down) is performed using bidirectional LSTM networks that exploit intrinsic mode function (IMF) representation of inertial signals. Records with inertial signals (accelerometer and gyroscope) of 2.56 s, available at the UCI Machine Learning Repository, were collected from 30 subjects using a smartphone. First, inertial signals were standardized to take them to the same scale and were decomposed into IMF using the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN). IMF were then segmented (split) into nine segments of 1.28 s with 12.5% overlap and introduced to a first network with four outputs to identify the dynamic activities and the statics as a single class called ‘‘statics’’, giving 98.86% accuracy. Then, the non-segmented IMF of the records assigned to the statics class were introduced to a second network to classify their three activities, giving an accuracy of 88.46%. In total, 92.91% accuracy was obtained to classify the six human activities. This performance is because ICEEMDAN allowed the extraction of information that was embedded in the signal, and the segmentation of the IMF allowed the network to discriminate between static and dynamic activities.
3
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Human activity recognition (HAR) from wearable motion sensor data is a promising research field due to its applications in healthcare, athletics, lifestyle monitoring, and computer–human interaction. Smartphones are an obvious platform for the deployment of HAR algorithms. This paper provides an overview of the state-of-the-art when it comes to the following aspects: relevant signals, data capture and preprocessing, ways to deal with unknown on-body locations and orientations, selecting the right features, activity models and classifiers, metrics for quantifying activity execution, and ways to evaluate usability of a HAR system. The survey covers detection of repetitive activities, postures, falls, and inactivity.
A study of data preprocessing influence on accelerometer-based human activity recognition algorithms is presented. The frequency band used to filter-out the accelerometer signals and the number of accelerometers involved were considered in terms of their influence on the recognition accuracy. In the tests four methods of classification were used: support vector machine, decision trees, neural network, k-nearest neighbor.
PL
W artykule przedstawiono wpływ przetwarzania wstępnego sygnału przyspieszenia na skuteczność rozpoznawania aktywności ruchowych. Przeanalizowano zależność filtracji sygnałów oraz ilości zastosowanych czujników na skuteczność klasyfikacji. W badaniach wykorzystano cztery różne klasyfikatory: maszynę wektorów wsparcia, drzewa decyzyjne, sztuczne sieci neuronowe oraz klasyfikator najbliższego sąsiada.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.