Project description

The very nature of surviving as a sentient being in a living world demands the ability to detect and differentiate other living entities around you, and to understand their behaviour and its potential effect on you. In other words, we use our sense systems to classify objects in the world around us.


Sound is the primary medium for long distance passive and active interaction between animals, and between some animals and their environment; ranging from human speech communication to the active acoustic scene analysis of bats and dolphins using bio-sonar. In this project we will try to emulate some of the ways in which animals use sounds to understand their world.

Aims

SCANDLE will develop a ground-breaking acoustic scene analysis system, ready for use in real-world environments and suitable for deployment in situations where visual information may be unavailable, unobtainable or even undesirable.

The system will be able to form representations of living beings and their behaviour exclusively through the use of information derived from sounds. The technological innovations will be firmly rooted in the latest scientific understanding of cognition in natural systems.

The proposed active/passive acoustic scene analysis system will be realised in real-time neuromorphic spike-based technology. The system we are developing is unique and will have the potential to support many new applications.

Applications

 

There are many real-world applications that can benefit from the technological innovations of this project. Examples include –

systems that can transparently and non-invasively provide situation awareness (warning signals) such as might be useful at busy road crossings
 Zebra Crossing Ginza Tokyo

 

detecting movement in a smoke-filled room where cameras are useless
 Truck on Fire by John Leach

 

detection of intentional human movements aimed at harming individuals or a group
 Shadows by Sub-Urban

 

recognition of normal/abnormal behaviour in a crowd
 Too Many People at Rock am Ring by rp72
Document Actions