Irek Defée

DCT histogram optimization for image database retrieval

Information extraction from images and video has been traditionally done in the pixel domain. Currently great majority of pictures and video are available in compressed form with compression based on block DCT transform. Compression removes significant amount of information leaving only perceptually important part and this has potential advantage from the information retrieval point. Optimization of compression for retrieval purposes is thus of interest but topic has not been much emphasized in the past. In this paper we study the problem of image database retrieval from the compression perspective. The approach is based on histograms of quantized DCT blocks. We show how these histograms can be optimized in order to achieve best retrieval performance by optimizing the selection of quantization factor and the number of DCT blocks under normalization of luminance. Results of experiments on face databases show that optimized histograms are robust in retrieval tasks. This indicates that selection and local feature compression optimization is an important step for effective pattern retrieval.

A wireless ballistocardiographic chair.

This paper presents a wireless ballistocardiographic chair developed for the Proactive Health Monitoring project in the Institute of Signal Processing. EMFi sensors are used for BCG measurement and IEEE 802.15.4 RF link for radio communication between the chair and a PC. The chair measures two BCG signals from the seat and the backrest and a rough ECG signal from the armrests of the chair. The R-spike of the ECG signal can be used as a synchronisation point to extract individual BCG cardiac cycles. Also, two developed methods for extracting BCG cycles without using a reference ECG signal are presented and compared

XMPP-based wireless sensor network and its integration into the extended home environment

Wireless sensor networks have significant potential in the home environment. Home application scenarios include monitoring and interaction from several remote or local devices such as mobile phones, televisions or computers. In this paper we propose an architecture for a network based on IP-enabled wireless sensors. The communication system is based on the XMPP-protocol which is shown to be highly suitable for the sensor networks. XMPP provides a push based notification functionality that is especially appropriate for the event-driven paradigm of sensors and also provides the authentication and encryption required for data protection. We also use Atom feeds together with XMPP to enable remote web service-like access to the wireless sensor network. Integration of this XMPP-based wireless sensor network with the standard UPnP media-oriented network is done using a XMPP to UPnP gateway. In this paper, home applications and services are first introduced and different technologies supporting them are described. Next, the benefits of XMPP/IP architecture compared to non-IP sensor networks and to the UPnP based sensor network are discussed. We then present our XMPPbased wireless sensor network, with web-service like access, using XMPP and Atom feeds, also interfacing with UPnP. These solutions have been implemented on internet tablet devices allowing a flexible, highly functional and open system for integrating sensor network with the media environment.

Performance of similarity measures based on histograms of local image feature vectors

We investigate similarity measures for image retrieval from databases based on histograms of local feature vectors. The feature vectors are obtained from grouping quantized block transforms coefficients and thresholding. After preliminaries on block transforms we are introducing binary DC and AC feature vectors. Subsequently ternary DC and AC vectors are defined. Next we show how the histograms of vectors defined can be combined to form similarity measure for image retrieval from database. We formulate the database training and retrieval problem using the defined similarity measures. Performance results are shown using widely used FERET and ORL databases and the cumulative match score evaluation. We show that despite simplicity the proposed measures provide results which are on par with best results using other methods. This indicates that statistics based retrieval should not be underestimated comparing to structural methods.

Mobile TV content to home WLAN

The proliferation of mobile broadcast networks and home multimedia networks makes relevant considering a new level of digital convergence. One aspect is the integration of mobile content access to the home network. In this framework, a mobile device serves as a gateway for real time content rebroadcast. This paper investigates the mobile rebroadcasting problem and implementation between the Mobile TV system based on the DVB-H standard and DLNA/UPnP enabled home network based on WLAN.

UUTE Home Network for Wireless Health Monitoring

This paper presents a home sensor network for wireless health monitoring, including a wireless sensor network, client for controlling the sensor network, and a data storage server. A common software and hardware microcontroller-sensor interface was defined to enable joint use of sensor technologies developed in three different projects. IEEE 802.15.4 RF-transceiver based radio-boards and ZigBee network software were designed and built, along with a simple sensor network software on top of the ZigBee stack, to implement the wireless sensor network. Both commercial and custom made sensors have been interfaced to the sensor network. A set-up consisting of four sensors was developed and tested in a real home environment. The architectural overview of the system and main technical design choices are presented.

Face Retrieval Based on Robust Local Features and Statistical-Structural Learning Approach

A framework for the unification of statistical and structural information for pattern retrieval based on local feature sets is pre-sented. We use local features constructed from coefficients of quantized block transforms borrowed from video compression which robustly preserving perceptual information under quantization. We then describe statistical information of patterns by histograms of the local features treated as vectors and similarity measure. We show how a pattern retrieval system based on the feature histograms can be optimized in a training process for the best performance. Next, we incorporate structural information description for patterns by considering decomposition of patterns into subareas and considering their feature histograms and their combinations by vectors and similarity measure for retrieval. This description of patterns allows flexible varying of the amount of statistical and structural information; it can also be used with training process to optimize the retrieval performance. The novelty of the presented method is in the integration of information contributed by local features, by statistics of feature distribution, and by controlled inclusion of structural information which are combined into a retrieval system whose parameters at all levels can be adjusted by training which selects contribution of each type of information best for the overall retrieval performance. The pro-posed framework is investigated in experiments using face databases for which standardized test sets and evaluation procedures exist. Results obtained are compared to other methods and shown to be better than for most other approaches.

A Non-beaconing ZigBee Network Implementation and Performance Study

In this paper we present our implementation of a layered ZigBee network and the performance test results using non-beaconing networks. Adding the implemented network layer to an existing IEEE 802.15.4 stack did not affect the throughput in the network. It is also concluded that with knowledge of the network structure an extra added delay between sending packets can reduce the number of packets lost due to channel access failure without decrease in throughput.

Notification service for DVB-H mobile broadcast

Notification is a service that addresses devices and users with messages to be processed immediately or at a specific time. Notification is important for delivering emergency, warning, and other timely messages. In this article extensions to include notification service in the data transport system called IP DataCast used in the DVB-H mobile television standard are presented. The architecture of IPDC in DVB-H is first overviewed, and the notification service elements complementing it are described. Next, the notification protocols and message formats are defined, and the system operation is explained. The presented notification system is now part of the new IPDC over DVB-H standard.

Utilization of wireless sensor network for health monitoring in home environment

The health care costs in developed countries are increasing fast due to the aging of the population. In-home monitoring of health is becoming more and more attractive both because of expected cost-savings and technical development of suitable measurement devices and wireless sensor networks. In this paper, we present on-going work about embedding health monitoring devices into ordinary homes. The developed system is targeted both for monitoring elderly and for monitoring rehabilitation after hospitalization period. The paper presents the utilized sensor network implementation, chosen set of sensors for the first test trial, as well as other design choices for the trial. In addition, further objectives about concentrating on one special case, the ubiquitous heart rate measurement, are discussed. Our objective is to install several non-contact heart rate monitors into a home environment. The designed system performed well during the trial. However, some issues, such as sensor addressing in WSN and user identification, will be better taken into account in the next trials.