Acticity Recognition Team
|
| Research Issues |
Since activity recognition is a relatively new field, there is lots of interesting research issues out there that still need attention. A few of them are following:
- Activity Classification: Recognition and classification of an activity is an important task of an activity recognition system. A single activity is usually a sequence of different sub-activities. E.g. “Going out for Work”, is a high level activity that can be pictured as a sequence of following sub-activities. Skipping any of these sub-activities, performing an additional sub-activity, that’s not the part of the sub-activity model or changing the order of sub-activities can result in the wrong classification of the high level activity.
|
|
- Activity Prediction: Most of the activity recognition systems identify the activity after it has been performed. But the goal of the ubiquitous systems is to understand the user preferences and take actions in advance. This requires automatic prediction of the user activity so that appropriate actions can be taken.
- Understanding the Context: Humans perform same activities but in different contexts which change the meaning of those activities form them. Thus considering only the activity performed without taking the context into account results in wrong decisions.
- Individual’s behavior: The behavior of an individual can be characterized by the temporal distribution of his activities such as patterns in timing, duration, frequency, sequential order, and other factors such as location, cultural habits, and age. These attributes also present challenges for activity recognition.
|
|
|
|
