Other Research Areas
|
Multimedia Systems and Sensor Networks
Multimedia systems involve the acquisition, format conversion, storage, transmission, processing and consumption of various forms of information. These forms typically involve text, audio, still images, animation, video and other forms of digital media. Multimedia systems are found in a variety of applications including entertainment and fine arts, journalism, engineering and medicine in order to effectively display and communication complex concepts. They are also emerging in advanced surveillance and healthcare (e.g., assisted living) applications.
Sensor networks proposed for surveillance, safety, and monitoring applications, are a class of information system that aims to change the way in which people observe and interact with their environment. Traditionally, sensor networks have been defined as being comprised of low-cost unattended groups of densely placed sensor nodes’ that observe, communicate (often using wireless means), and coordinate to collectively achieve high-level inference tasks. These networks ideally have the following characteristics; they are:
- distributed, in order to improve the performance and geographical range of sensing functions,
- data-centric, in which network processing is dependent on the sensed data instead of the particular identity of the node at which it was observed,
- collaborative, making use of the coordination of localized algorithms to achieve a global task with better scalability,
- redundant, using distributed densely deployed sensor nodes to obtain more accurate and complete readings of observed events,
- autonomous, for adaptability to change in network topology, and fault-tolerance without requiring servicing over long periods of time,
- application-specific, requiring in-network processing, such as data aggregation (a.k.a. fusion), to produce meaningful data about the observation area,
- hierarchical, through the localized clustering of sensor nodes into “sub-networks” to improve network scalability, and
- resource-constrained, necessitating the sparing use of communication bandwidth, memory and computation to reduce exhaustion of the often portable power source, which is one of the biggest design challenges in existing literature.
When the sensor nodes collect diverse types of information such as temperature, humidity, acoustic and visual data simultaenously, they are termed “multimodal sensors.” Multiple types of sensing can occur within the same node through the use of distinct sensing technologies or across different nodes each having a single, but distinct sensor type. Multimodal sensors that collect multimedia information such as digital images, video and audio, form a multimedia sensor network. The proliferation of advanced surveillance systems and modern personal devices (cell phones, tablets) with multimedia acquisition capability has popularized research in multimedia sensor networks.
Research Focus
Our work in multimedia sensor networks has four main thrusts. The first is on security and privacy-enhanced architectures. Game and graph-theoretical formulations are employed to understand design principles for multimedia sensor network systems. Our second thrust involves identifying network vulnerabilities and proposing security mechanisms for directional sensor networks capable of carrying multimedia information. Here we determine fundamental trade-offs between system connectivity and security for directional sensor networks based on random scaled sector graphical models. Third, we develop lightweight and distributed multimedia enciphering approaches suitable for sensor networks. Most recently we have identified characteristics of multimedia sensor systems that can promote and discourage steganography.