ABSTRACT
A sense of the past is useful because context does not change completely randomly, but instead shows some coherence over time. Although an isolated sensor measurement might lead to an uncertain conclusion about a person’s context, repeated measurements give more certainty in spite of noisy measurements, partly because context normally does not change
very quickly compared to how oen measurements can be taken. us, it oen makes sense to average together the last few measurements, which is a simple means of exploiting continuity, and one which this chapter examines in Section 9.3. In general, a new measurement triggers a reestimation of the context state variables that is sensitive not only to the new measurement, but also to previous measurements and/or estimates. is historical perspective means the techniques can exploit expectations about the continuity of the state variables. e mean lter fullls expectations that the output varies relatively slowly and that it should be sensitive to current and past measurements. All the techniques in this chapter pay attention to past measurements or past state estimates in some way.
