Fusing Multiple Sources of Context Data of the Same Context Type

Smart homes are expected to have sensors that monitor the home and its inhabitants. Sensors types include temperature sensors, pressure sensors in the ,floor; or even accelerometers worn by the inhabitants. Data from these sensors can be used to infer higherlevel context, such as the current activity of the inhabitants. We look at the case where there are multiple providers of the same type of context in the home, and describe a probabilisric mechanism for combining these multiple context outputs into a single output using the advertised probability of correctness of each provider. We show that the resulting context value has a higher quality (specifically a higher probability of correctness) than selecting and using any of the available providers on their own.