SCANDLEProject reference: 231168
Funded under: FP7-ICT
acoustic SCene ANalysis for Detecting Living Entities
Total cost:EUR 3 478 915
EU contribution:EUR 2 649 938
Coordinated in:United Kingdom
Call for proposal:FP7-ICT-2007-3
Funding scheme:CP - Collaborative project (generic)
An acoustic system which can recognise human or animal movement.
Living beings constantly give sonic clues about their presence as they move, breathe and interact with the world around them. Sounds also bounce off other objects in the environment; a source of further contextual information. In SCANDLE we propose to develop a cognitive acoustic scene analysis system that will detect and characterise the behaviour of living beings solely through sounds generated or modulated by their actions and interactions with the environment; an acoustic analogy to a camera-based visual scene analysis system.
The ultimate goal of SCANDLE is to engineer a real-time system that uses sound alone to decide whether an animate being is nearby, where it is, and what it is doing. In doing so the system will also make use of a novel micro-Doppler sonar system to generate and detect modulations in sounds caused by movements. This highly innovative multidisciplinary project builds upon fundamental research by ourselves and others in the fields of physics, acoustics, auditory neuroscience, human psychophysics, engineering and computer science. In addition, we are conducting new experiments designed to answer specific scientific questions related to the neural correlates of object formation and representations in the brain.
In SCANDLE we draw on biological inspiration in the design of a computational system, emulating the perceptual processing strategies employed by humans and animals for robust real-time operation in real-world scenarios. By understanding the constraints and costs associated with design choices in the brain, SCANDLE is also making advances in optimising the design of distributed computing architectures for massively parallel computation. The net result is a methodology for determining the architecture of specialised embedded systems and for compiling neural models for real-time operation on these systems.
In coming years this technology will give rise to many innovative and valuable applications, moving us into the realm of truly neuromorphic cognitive machines. For example, the system could be used to intelligently monitor home environments to facilitate independent living for the elderly or to optimise energy usage; using only sound information, issues of the invasion of privacy are reduced. Interactive gaming systems that understand peoples’ behaviour, remote monitoring and identification of animals or fish based on their movements, and detecting life where visual contact is obscured are just some of the exciting applications that will become possible using SCANDLE technology.
Analysing fine motor activity in articulatory structures of humans or animals in combination with the sounds they emit yields information about their intentions and likely future actions. In this project we propose to develop a cognitive acoustic scene analysis system that is able to synthesize composite representations of animate entities and their behaviour by integrating information from active and passive sound signatures; i.e. from actively self-generated (sonar) sounds and from passively received sounds emitted by those entities. The system is an acoustic analogy to a camera-based visual scene analysis system, particularly suited to detecting the presence and characterising the behaviour of living entities in the environment. This highly innovative proposal builds upon fundamental research on perceptual organisation in natural systems, recent advances in models of auditory processing, and technological developments in ultra-low power, distributed neuromorphic systems and state-of-the-art micro-sonar technology. The biologically inspired architecture and processing mechanisms of the proposed system support autonomous real-time context-dependent operation, allowing it to parse complex mixtures of sounds into meaningful units and categorise them. We propose novel methodologies for evaluating the emergence of representations in autonomous systems, and for communicating the ongoing internal state of the system to human observers. Work on the project will significantly advance scientific understanding of auditory perceptual organisation, technological developments in neuromorphic systems, and the potential impact of artificial cognitive systems. Successful achievement of our objectives will result in the development of a ground breaking proof-of-concept cognitive acoustic scene analysis system capable of robust operation in real-world environments and suitable for deployment in situations where visual information may be unavailable, unobtainable or even undesirable.
PL4 8AA PLYMOUTH
Tel.: +41 44 6353024
Fax: +41 44 6353053
Tel.: +36 1 3542296
Fax: +36 1 3542416