|
|
|
| UvA | Fac. Science | IAS | Tim van Kasteren's Homepage | ||
|
There are a large number of applications that deal with the recognition of activity patterns of one or more users. These applications exist primarily in the field of security (e.g. aggression, theft, terrorism), health (health monitoring) and comfort (control of lights, heating, music, etc). All these areas together form the Centre for Intelligent Observation Systems (CIOS) In the CARE project we study human activity patterns by using a large number of 'simple' sensors (e.g.contact switches, pressure mats, etc). The main focus in this project lies on health monitoring elders in their private homes, allowing them to live independently at home. The use of simple sensors minimizes the feeling of privacy breach and allows people to live their lives as they did without the system. We have formed a fruitful collaboration with several educational parties within the CARE project. At the University of Amsterdam (UvA) we research probabilistic methods for performing activity recognition from simple sensors. Students from the Institute Information Engineering (IIE) are studying the best way to implement a sensor network in an actual healthcare enviroment. Finally, we have students at the HvA school of nursing who provide us with information from the healthcare domain. Please check back here regularly for a progress update of the project. The CARE project is partly funded by the Centre for Intelligent Observation Systems (CIOS) which is a collaboration between UvA and TNO, and partly by the EU Integrated Project COGNIRON: The Cognitive Robot Companion. | 
|
|
The tempo in music is the speed at which a piece of music is played. A metronome is a mechanical device which can be used to indicate the tempo by providing clicks at a constant pace (see picture right). When a musician uses a metronome while playing his instrument, he carefully listens to the clicks of the metronome to determine the tempo at which he should play. In tempo tracking we try to reverse this process, that is, a musician can play at any tempo he wishes and the computer will listen to the musician to determine the tempo and play the clicks at the appropriate time. What makes tempo tracking so difficult, is that the computer does not know in advance what the musician will play. Because the tempo is also unknown each observation is very ambiguous. An observed note can be interpreted as a quarter note played at a particular tempo or as an eighth note played at half that tempo. A lot of other combinations of tempo and note duration are possible at each observation. And because music contains all sorts variations in tempo it is difficult to come up with a consistent interpretation. In the thesis a method is described which is capable of handling these difficulties in realtime. By using a Kalman filter minor fluctuations in tempo (intentionally or unintentionally) are filtered out. Through the use of a particle filter various interpretations are tested to determine the most likely. |
|
