In ubiquitous home network, recognition of daily activities (Taking medication, sleeping, etc.) is one of the current focuses of Researchers. Especially in health care industry it has a huge impact. As an example, monitoring daily activities can reduce the risk of elderly people or chronically ill children.
Three different approaches have been tried by the researchers to detect human activity: video based, wearable sensors based, and based on sensors deployed in the environment. In these approaches, only limited number of activities was focused. And the accuracy is not up to the mark. On the other hand we will recognize thousands human activities from micro to macro using the combination of all these approaches such that a better accuracy can be achieved.
Unavailability of data set that represents activities makes it difficult to progress in activity recognition research. Real world data is hard to find. Even if we can obtain this, there will be noises. Existing deployments, generates thousands of sensor data from the environment but very few activities are instantiated. Collecting the data and help the research community is also a part of our research. |
