Fast prototyping of a SoC-based smart-camera: a real-time fall detection case study
B. Senouci, ︎ I. Charfi, ︎ B. Heyrman ︎, J. Dubois ︎, J. Miteran
Smart camera, i.e. cameras that are able to acquire and process images in real-time, is a typical example of the new embedded computer vision systems. A key example of application is automatic fall detection, which can be useful for helping elderly people in daily life. We propose a methodology for development and fast-prototyping of a fall detection system based on such a smart camera, which allows to reduce the development time compared to standard approaches.
Founded on a supervised classification approach, we propose a HW/SW implementation to detect falls in a home environment using a single camera and an optimized descriptor adapted to realtime tasks. This heterogeneous implementation is based on Xilinx’s system-on-chip named Zynq.
The main contributions are (i) the proposal of a co- design methodology. These methodologies enable the HW/ SW partitioning to be delayed using high-level algorithmic description and highlevel synthesis tools. Our approach enables fast prototyping which allows fast architecture exploration and optimisation to be performed, (ii) the design of a hardware accelerator dedicated to boosting- based classification, which is a very popular and efficient algorithm used in image analysis, (iii) the proposal of fall- detection embedded in a smart camera and enabling integration into the elderly people environment. Performances of our system are finally compared to the stateof-the-art.
Keywords : Real-/me fall detec/on , SoC implementa/on , Fast smart camera prototyping , Zynq , HW/SW
implementa/on , Boos/ng hardware implementa/on
Le2i UMR 6306, University of Burgundy,
Contact : J. Dubois | julien.dubois@u-bourgogne.fr
Security and self-organizing mechanisms for the Internet of Things
Axel Moinet & Benoit Darties
The last ten years have seen the development of mobile technologies, from smartphones to everyday life connected devices and cloud computing. In this context, Wireless Sensor Networks (WSN) have emerged from specific industrial applications to everyone. These networks encounter rout- ing and organization issues, especially in the context of the IoT, where needs and resources between devices are different. This project introduce the experimentation testbed we are deploying to check routing and topology correctness of simulations and research work, using the 802.11s mesh net- work protocol as a base.
Networking technologies have grown to many sub-domains, from Cloud computing to Wireless Sensor Networks (WSN) and even social networks relying persons. The “Future Internet” is the main term to refer to all interconnections between these networks. In these interconnections, the IoT is the branch which integrates WSNs to end user cloud-based applications. A middleware is used between the cloud and WSNs, providing the application with information about and interaction with the environment. It can be a particular environment, like a house or office, or a more global environment, as an entire country.
We focus our research on self-organization and routing inside WSNs, the root part of the IoT. Globally, network dimensioning is a recurrent problem in WSN, as node number and evolutivity of the network topology causes routing and network map construction issues.
Wireless Sensor Networks are build up by low power and low resources heterogenous devices. Since these devices have different purpose, they also have different material architecture and resources levels. Energy and computing re- sources are a main concern in this field, because sensors don’t have a fixed power supply, making batteries the energy supplier for all device lifetime. Moreover, devices may be mobile, and they may become defective [2]. The main principle is to allow devices to spread in the net- work when they need. This way, each device energy saving and resource control policies will be able to manage connection time, to save power and/or computing time. This also resolves the mobility and defectively issues, by making no node absolutely necessary in the network. Thus, the organization of devices inside the network must be evolutive, but still secure [2]. The most efficient way of building mutable networks is to use a mesh networking infrastructure, which provides connection with neighbouring nodes. Since some theoritical and simulation research work have already been made in this domain, we choose an experimental approach.
2.1. Testbed
We currently are deploying a testbed aimed at testing multi-hop communications in a network composed of heterogeneous nodes, in real conditions. The short-term objective is to have an experimentation platform, allowing us to exceed simulators limits. This will also help us to characterize security, energy and computing costs for different strategies and topologies, in real conditions. We choose 802.11s as the mesh protocol inside our network, because it as many advantages. First, it allows us to easily modify or implement our own routing protocol and strategies. Second, it provides standardized high-security connectivity mechanisms [1]. And it also has the advantage to be implemented in the Linux kernel since the IEEE ratified the standard.
Our testbed will be composed of four different devices with different architectures and resources :
ARM Cortex- A8, ARM Cortex-A5, ARMv6 1176jzf-s and ARM Cortex- A7.
Contact : B. Darties | benoit.darties@u-bourgogne.fr
Priority Levels Based Multi-hop Broadcasting Method for Vehicular Ad hoc Networks
Wahabou Abdou · Benoît Darties · Nader Mbarek
A Vehicular Ad hoc Network (VANET) is a collection of vehicles communicating through wireless connections: each vehicle acts simultaneously as a node and as a wireless router, allowing multihop packet forwarding. Indeed, each node has a limited coverage area that contains the neighbors it can directly communicate with. This area can vary from one hundred meters to a few kilometers. This project deals with broadcasting techniques which are used for sending safety messages,
traffic information or comfort messages. When a packet is broadcasted, it is received by all nodes within the sender’s coverage area (provided that no interference or radio channel trouble occurs). Every receiver will decide to relay or not the packet depending on its own broadcasting strategy.
This hop-to-hop communication would lead to a full coverage of the network. Performing an efficient multi-hop broadcast in VANETs is however a difficult task. The protocols should take into account the specificities of the radio channel, the high mobility of nodes and the network density. The decision to relay the packets is taken in a distributed way, but each node’s decision has a real impact on the efficiency of the overall dissemination strategy: in high-density networks, too many relay nodes would quickly increase the number of collisions, leading to a saturation of the bandwidth and a significant increase of the latency. On the other hand, if not enough relay tasks are performed in low-density networks the message may not be widely disseminated. Several approaches are proposed in the literature to overcome this problem. Some methods from Mobile Ad hoc Networks (MANETs) use the neighborhood knowledge to choose the best relay nodes. However, the high mobility that characterizes VANETs makes their application in such a network very difficult. Some stochastic methods use waiting time to reduce the number of redundant packets. But if this waiting time is not well chosen, it may increase the message dissemination time. The approach we present in this paper is based on the Smart-flooding protocol that uses a genetic algorithm optimization process to dynamically adapt dissemination strategies with respect
to the network density.
It is also important to adapt the broadcasting strategy to the priority level of the message. For instance, emergency messages such accident alerts, should be delivered as fast as can be done in the source node’s neighborhood. Conversely, it does not matter if some weather information with limited impact and less urgent tourist information are broadcasted with a more important latency since there is no emergency.
We started investigating the problem of building an autonomous and robust broadcasting protocol, which provides each node with the adequate strategy to determine if an incoming message has to be forwarded or not depending on its priority level and the network density. The goal was to make effective use of radio resources when there are many messages to send simultaneously. In this project we expand the study of the applicability of this novel approach to various types of networks
(in terms of density level).
Contact : Wahabou Abdou | wahabou.abdou@u-bourgogne.fr
Le2I – Université de Bourgogne - More items from the Field#1 :
