Fujitsu Laboratories has developed a new system that would allow a regular tablet camera or PC webcam to accurately capture and recognize three dimensional motions and gestures.
Fujitsu Laboratories have just announced last May 20, 2013 the development of their new motion recognition system. With the use of a PC webcam, or a standard camera mostly used on mobile devices, they are able to make the new system detect and recognize motion and gestures in three dimensions.
Fujitsu's new system is based on two fundamental techniques. The first one is the development of a block model of the object. They first build image-based boundaries of the space that's supposed to be occupied by the object. The boundaries are then used to create a "minced" digital version of the object that is made up of different tiny digital blocks (pixels?). By analyzing the relative change in the size and number of digital blocks seen on screen, the system could then calculate its depth, therefore providing a measurement of how far the object is from the camera.
The second technique involves the straightforward method of simply analyzing the change of the object's size in comparison to the static surroundings. This also includes the time and rate, or how fast the change in size occurred. This would of course be used to measure just how far the object actually is from the camera.
The combination of the two measurements allows their system to accurately use a "push" or "pull" command (or a click gesture) whenever relative distance measurement is required. Aside from counting digital blocks and measuring the default size of the object, the system also scans for other variables, such as the object's color and background hue in order to make the analysis more accurate.
Standard single-camera based hand gesture recognition systems that we have today often only work in two dimensions: up and down, left and right. This is quantified by the motion space occupied by a cursor on a screen. Because of this dimension limitation, there were often difficulties in finding an easy and intuitive gesture for initiating a "click" command. By converging two different simple but accurate depth recognition methods, Fujitsu was able to make their 3D gesture recognition system easy and flexible to use. In addition, because their system doesn't use a package of different sensors and equipment just like one very popular device out there, high costs (for application and use) can be significantly mitigated.
Source: MyNavi (JP)