Heres a few more notes on the papers I read yesterday. The first one is titled "Inferring Activities from Interactions with Objects" its the Intel research Seattle project. It begins with discussing the difficulty of using computers and sensors for interpreting and recognising the activities of individuals as they indertake everyday tasks, and more specifically in the care of elderly people. The main problem seems to be getting the recognition software to adapt to "variety", for example the same task could be completed in different ways. Their solution to this is to represents tasks in a sequence or probable sequence of the objects used.
The basic system is the same as my own project where objects are RFID tagged and read from a mobile RFID reader, in this case an antenna embedded in the palm of a glove. The software implementation is based on probabilistically estimating activities from the sensor data, taking into account different variables such as time. For example in the given scenario, in the frist step of making a cup of tea there will be a high probability of encountering a kettle. This seems like a logical approach to the problem. The system was tested with people undertakning a number of everyday tasks. My own project will need to be tested in a similar way. The results varied depending on the task but were promising overall. One thing thats important to note from the results is that water and metal are bad! Water absorbs the radio waves and metal could short circuit the antenna.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment