Activity Recognizition Teams
Team name
Chiba Yu-Lab
|
|
Corresponding author Nevrez Imamoglu
|
Affiliation Japan Chiba University
|
Description
Our previous studies demonstrated that the idea of bio-monitoring home healthcare mobile robots is feasible. Therefore, by developing algorithms for mobile robot based tracking, measuring, and activity recognition of human subjects, we would be able to help impaired people (MIPs) to spend more time focusing in their motor function rehabilitation process from their homes. In this study, we aimed at improving two important modules of these kind of systems: the control of the robot and visual tracking of the human subject.
For this purpose: 1) tracking strategies for different types of home environments were tested in a simulator to investigate the effect on robot behavior; 2) a multi-channel saliency fusion model with high perceptual quality was proposed and integrated into RGB-D based visual tracking. Regarding the control strategies, results showed that, depending on different types of room environment, different tracking strategies should be employed. For the visual tracking, the proposed saliency fusion model yielded good results by improving the saliency output. Also, the integration of this saliency model resulted in better performance of RGB-D based visual tracking application.
|
Team name
JSI |
|
Corresponding author Simon Kozina
|
Affiliation Jožef Stefan Institute (SI)
|
Description Ambient assisted living (AAL) systems need to understand the user’s situation, which makes activity recognition an important component. Falls are one of the most critical problems of the elderly, so AAL systems often incorporate fall detection. Our activity recognition (AR) and fall detection (FD) system aim to provide robust real-time performance. It uses two wearable accelerometers. For the AR, we developed an architecture that combines rules to recognize postures, which ensures that the behavior of the system is predictable and robust, and classifiers trained with machine learning algorithms, which provide maximum accuracy in the cases that cannot be handled by the rules. For the FD, rules are used that take into account high accelerations associated with falls and the recognized horizontal orientation (e.g., falling is often followed by lying).
|
Team name
AReM |
|
Corresponding author Filippo Palumbo
|
Affiliation CNR - ISTI (IT)
|
Description An activity recognition system that classifies a set of common daily activities exploiting both the data sampled by accelerometer sensors carried out by the user, and the Receive Signal Strength (RSS) values coming from wireless sensors devices deployed in the environment. To this end, accelerometer and RSS streams, obtained from a Wireless Sensor Network (WSN), are treated using Recurrent Neural Networks implemented as efficient Echo State Networks (ESNs), within the Reservoir Computing paradigm.
|
Team name
AmevaActivity
|
|
Corresponding author Miguel Ángel Álvarez de la Concepción
|
Affiliation University of Seville (ES)
|
Description AmevaActivity aims to develop a cheap, comfortable and, specially, efficient system which controls the physical activity carried out by the user. For this purpose an extended approach to physical activity recognition is presented, based on the use of discrete variables which employ data from accelerometer sensors. To this end, an innovative selection, discretization and classification technique to make the recognition process in an efficient way and at low energy cost, is presented in this work based on Ameva discretization. Entire process is executed on the smartphone and on a wireless health monitoring system is used when the smartphone is not used taking into account the system energy consumption.
|
Team name
KinectHAR
|
|
Corresponding author Georgiana Scarlat
|
Affiliation University POLITEHNICA of Bucharest, Computer Science Department, Bucharest, Romania (RO)
|
Description The system recognizes activities of daily living using information provided by four Kinects. First the posture of the supervised person is detected using a set of rules created with ID3 algorithm applied to a skeleton obtained by merging the skeletons provided by multiple Kinects. At the same time, the interaction of the user with the objects from the house is determined. After that, daily activities are identified using Hidden Markov Models in which the detected postures and the object interactions are observable states. The benefit of merging the information received from multiple Kinects together with the detection of the interaction between the user and relevant objects from the room increase the accuracy of the recognized activities.
|
.
|