Cookies Policy
We use cookies to improve our site and your experience. By continuing to browse our site you accept our cookie policy. Find out More
Close
  • Menu
About

About

Maria Teresa Andrade is an Assistant Professor at FEUP, at DEEC. She obtained a degree in Electrotechnical and Computing Engineering in 1986, the MSc in 1992 and the PhD in 2008, at FEUP. She participates in research activities at INESC TEC, integrated in the research team of the Multimedia Systems Area of the Center for Telecommunications and Multimedia. Main interests include context-awareness, mobile and adaptable multimedia applications in heterogeneous environments; 3D and multiview video streaming; quality of service and of experience in multimedia services; semantic technologies and content recommendation; digital television, digital cinema and new media.

Interest
Topics
Details

Details

011
Publications

2018

GymApp: A real time physical activity trainner on wearable devices

Authors
Viana, P; Ferreira, T; Castro, L; Soares, M; Pinto, JP; Andrade, T; Carvalho, P;

Publication
Proceedings - 2018 11th International Conference on Human System Interaction, HSI 2018

Abstract
Technological advances are pushing into the mass market innovative wearable devices featuring increasing processing and sensing capacity, non-intrusiveness and ubiquitous use. Sensors built-in those devices, enable acquiring different types of data and by taking advantage of the available processing power, it is possible to run intelligent applications that process the sensed data to offer added-value to the user in multiple domains. Although not new to the modern society, it is unquestionable that the present exercise boom is rapidly spreading across all age groups. However, in a great majority of cases, people perform their physical activity on their own, either due to time or budget constraints and may easily get discouraged if they do not see results or perform exercises inadequately. This paper presents an application, running on a wearable device, aiming at operating as a personal trainer that validates a set of proposed exercises in a sports session. The developed solution uses inertial sensors of an Android Wear smartwatch and, based on a set of pattern recognition algorithms, detects the rate of success in the execution of a planned workout. The fact that all processing can be executed on the device is a differentiator factor to other existing solutions. © 2018 IEEE.

2018

Multimedia Content Classification Metrics for Content Adaptation

Authors
Fernandes, R; Andrade, MT;

Publication
U.Porto Journal of Engineering

Abstract
Multimedia content consumption is very popular nowadays. However, not every content can be consumed in its original format: the combination of content, transport and access networks, consumption device and usage environment characteristics may all pose restrictions to that purpose. One way to provide the best possible quality to the user is to adapt the content according to these restrictions as well as user preferences. This adaptation stage can be best executed if knowledge about the content is known a-priori. In order to provide this knowledge we classify the content based on metrics to define its temporal and spatial complexity. The temporal complexity classification is based on the Motion Vectors of the predictive encoded frames and on the difference between frames. The spatial complexity classification is based on different implementations of an edge detection algorithm and an image activity measure.

2017

Context-Aware Personalization Using Neighborhood-Based Context Similarity

Authors
Otebolaku, AM; Andrade, MT;

Publication
Wireless Personal Communications

Abstract
With the overwhelming volume of online multimedia content and increasing ubiquity of Internet-enabled mobile devices, pervasive use of the Web for content sharing and consumption has become our everyday routines. Consequently, people seeking online access to content of interest are becoming more and more frustrated. Thus, deciding which content to consume among the deluge of available alternatives becomes increasingly difficult. Context-aware personalization, having the capability to predict user’s contextual preferences, has been proposed as an effective solution. However, some existing personalized systems, especially those based on collaborative filtering, rely on rating information explicitly obtained from users in consumption contexts. Therefore, these systems suffer from the so-called cold-start problem that occurs as a result of personalization systems’ lack of adequate knowledge of either a new user’s preferences or of a new item rating information. This happens because these new items and users have not received or provided adequate rating information respectively. In this paper, we present an analysis and design of a context-aware personalized system capable of minimizing new user cold-start problem in a mobile multimedia consumption scenario. The article emphasizes the importance of similarity between contexts of consumption based on the traditional k-nearest neighbor algorithm using Pearson Correlation model. Experimental validation, with respect to quality of personalized recommendations and user satisfaction in both contextual and non-contextual scenarios, shows that the proposed system can mitigate the effect of user-based cold-start problem. © 2016 Springer Science+Business Media New York

2017

2D/3D Video Content Adaptation Decision Engine Based on Content Classification and User Assessment

Authors
Fernandes, R; Andrade, MT;

Publication
PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON NUMERICAL ANALYSIS AND APPLIED MATHEMATICS 2016 (ICNAAM-2016)

Abstract
Multimedia adaptation depends on several factors, such as the content itself, the consumption device and its characteristics, the transport and access networks and the user. An adaptation decision engine, in order to provide the best possible Quality of Experience to a user, needs to have information about all variables that may influence its decision. For the aforementioned factors, we implement content classification, define device classes, consider limited bandwidth scenarios and categorize user preferences based on a subjective quality evaluation test. The results of these actions generate vital information to pass to the adaptation decision engine so that its operation may provide the indication of the most suitable adaptation to perform that delivers the best possible outcome for the user under the existing constraints.

2016

User context recognition using smartphone sensors and classification models

Authors
Otebolaku, AM; Andrade, MT;

Publication
JOURNAL OF NETWORK AND COMPUTER APPLICATIONS

Abstract
Context recognition is an indispensable functionality of context-aware applications that deals with automatic determination and inference of contextual information from a set of observations captured by sensors. It enables developing applications that can respond and adapt to user's situations. Thus much attention has been paid to developing innovative context recognition capabilities into context-aware systems. However, some existing studies rely on wearable sensors for context recognition and this practice has limited the incorporation of contexts into practical applications. Additionally, contexts are usually provided as low-level data, which are not suitable for more advanced mobile applications. This article explores and evaluates the use of smartphone's built-in sensors and classification algorithms for context recognition. To realize this goal, labeled sensor data were collected as training and test datasets from volunteers' smartphones while performing daily activities. Time series features were then extracted from the collected data, summarizing user's contexts with 50% overlapping slide windows. Context recognition is achieved by inducing a set of classifiers with the extracted features. Using cross validation, experimental results show that instance-based learners and decision trees are best suitable for smart phone -based context recognition, achieving over 90% recognition accuracy. Nevertheless, using leave one -subject-out validation, the performance drops to 79%. The results also show that smartphone's orientation and rotation data can be used to recognize user contexts. Furthermore, using data from multiple sensors, our results indicate improvement in context recognition performance between 1.5% and 5%. To demonstrate its applicability, the context recognition system has been incorporated into a mobile application to support context-aware personalized media recommendations.

Supervised
thesis

2017

Social TV: A integração entre a televisão convencional e as redes sociais

Author
Paulo António da Silva Brandão

Institution
UP-FEUP

2017

Interfaces Conversacionais – Chatbot para a Casa da Música

Author
Sara Filipa Gomes Oliveira

Institution
UP-FEUP

2017

Mobile TV

Author
Cristhian David Santo Caldeira

Institution
UP-FEUP

2017

Artistic Style Transfer for Textured 3D Models

Author
Inês Filipa Nunes Teixeira

Institution
UP-FEUP

2017

MATT - Media Asset Tracking Tool

Author
Tomás Fernandes Brandão Tavares

Institution
UP-FEUP