Kinect-based system keeps all your activities under observation

You might have experienced the Kinect technology for gaming until now, but it is now being used in a very different way. Everyone wants to live a very comfortable life, we even sometimes dream of a life when there will be robot assistants ready to lend us a hand all the time. Many people think that the best thing of having robot assistants is that, one can do anything in front of them, as they are just machines. If you think in a similar manner, then think twice, as things have changed a lot.

Kinect-enabled system

A Microsoft Kinect-based system has been developed, which can recognize what people are doing with the help of the motions they make. All this has become possible because of the research presently being carried out at Cornell University. Further, if this sort of technology is integrated in a robot, then a robot might warn you if you chew with your mouth wide open, or if you find it difficult to lift a heavy object, it can actually come forward to offer you help in lifting that object. This implies that you will have to be particular about the activities that you do in front of such robots.

This research project utilized the RGBD (Red, Green, Blue and Depth) camera from Kinect for observing four people, who executed twelve different actions in five separate settings. These settings included kitchen, office, living room, bedroom, and bathroom. The activities these people performed included cooking, working on a PC, drinking water, writing on the whiteboard, and even brushing teeth. After this, the data was dashed all the way through an algorithm, where these activities were classified into more convenient sub-activities. This is called the hierarchical maximum entropy Markov model, and it affirms that if any person is spotted to do activities A, B, C and then D, then what they are actually doing is probably E.

Working of the system

When this system observes a person carrying out the activities in training set, then it recognizes those actions with an accuracy of 84.3% ‘in the field.’ In case, the system has not observed the person earlier, then its accuracy of recognition was 64.2%, which is obviously less than the first instance.

Kinect-enabled system

Its future

If this technology is embedded in personal assistant robots, then they can be very helpful in household chores. It can check if people are drinking sufficient water, taking their medication on time, brushing their teeth on a regular basis, and so on. During the research, this system steered clear from being thrown off by the unrelated gestures, which were mixed along with the activities. The researchers agree that they are not completely certain about the performance of this system, if objects were obstructing its view. They also recommended that this system would work more efficiently if it could learn to identify some key objects like forks, or toothbrushes, which are related to a few activities.

Via: I-Programmer/Gizmag