Abstract:

Falling is a feature of human ambulation which is known to result in injuries, loss of mobility, hospitalisation, high medical expenses and even fatalities. Ambulation monitoring can be applied to individuals susceptible to falls in order to provide prompt medical assistance and a qualitative assessment of stability. Statistics show that the elderly are a demographic with a high susceptibility to falling and suffering from fall-related injuries. The report begins by introducing the theories and mechanisms behind human ambulation and Artificial Intelligence's dynamic belief networks (DBNs) - a probabilistic reasoning system which can be used to model and produce conclusions about the state of complex temporal environments. The remainder of the report discusses the implementation of a system utilising DBN theory to perform reasoning in the temporal and inaccessible domain of human walking, for the purpose of ambulatory status monitoring and fall detection. Discussion of the project's hardware covers a commercial datalogger employed to record and transmit ambulation data, investigation into applicability of existing sensors and the subsequent development of new pressure sensors. Software aspects discussed include the development of a configurable DBN-based data processing system, and a number of dynamic belief network models developed and used with this system to recognise different ambulatory situations. Results show the system successfully recognising phases of the gait cycle and various forms of stable and unstable ambulation, including stopping, walking, running, stumbling and falling.