Team: Ramona Rednic, Dr. Elena Gaura, John Kemp, Dr. James Brusey.
The project aims to enable the pursuit of new lines of clinical quantitative research for posture and mobility related medical conditions, through the development of wearable instrumentation. The instrumentation will allow both patients in need of diagnosis, monitoring or assessment and medical experts to gather and review posture and mobility patterns over selected periods of time. For patients, the self monitoring feature will be enabled through the delivery of the information on demand on PDAs and other mobile devices. For medical experts, remote monitoring and storage of the mobility/posture data and inferred information will be enabled wirelessly through web-posting. For both user groups, the key value added is through the “patterns in front of your eyes” concept which allows global interpretation of the subject’s actions and fully exploits one of the greatest features of wireless sensor networks: the ability to build phenomenologicaltemporal- spatial links allowing global observations to be drawn from data.
The project's objectives are:
It is well known and understood by both clinicians and wearable technology developers that early prototype evaluations and inclusion of the staged evaluations within the main design processes is essential. On the other hand, given the long term wearability needs for the product proposed here and also the information driven proposed designs, it is important to validate each step of the design process in un-obtrusively instrumented environments where the subjects can perform activities no different than the ones in their normal routines. This will allow a much wider range of data and observations to be accounted for in the information extraction processes performed by the proposed systems, increase the system reliability and correctness of decisions and also ensure user acceptability and deep understanding of the system’s aims. For these reasons the project is heavily experimental in approach.
A prototype wearable posture assessment system has been designed, implemented and tested. To date, the prototype is able to correctly classify 8 human postures, at a rate of 60 assessments per second and relay the posture information remotely via Bluetooth to the user. The posture assessment information is delivered in graphical form. The system uses 9 micromachined 3-axial acceleration sensors, connected for wearability to 2 processing and communication nodes. The hardware support for the nodes is the Gumstix platform. All computational tasks are supported within the network; data acquired is filtered, outliers are rejected and the posture information is inferred from the data. The posture inference is realized through the use of a decision tree. Currently the system has been evaluated as being able to classify correctly between the 8 postures in 96% of cases.
Go to Ramona's PhD project page for details on continuing work.