Gesture recognition in real time can bridge a gap between humans and computers. Object segmentation from the background is a critical problem in the conventional gesture

A natural man-machine interface requires gesture recognition. In gesture-recognition technologies, extraction of the target image (for example a hand or an arm) from its background images is the most important technology. Generally, gesture recognition requires 3D extraction at over 60frames/s. However, extraction frame rate has been slow.Here, real-time 3-Dimensional object extraction is by a recognition system which consists of a CMOS image sensor and infrared light emitting diodes (LED). The CMOS image sensor features two storage capacitors per pixel, on-chip column differential circuits, and 8b analog-to-digital converter (ADC). The extraction camera with LEDs and the CMOS sensor produces a high-quality 3Dimage using a 3.3V supply and consuming 50mW for the image sensor.Figure 7.7.1 shows a conceptual diagram of the 3-D imageextraction camera. It consists of the CMOS image sensor, LEDs and a lens. The sensor outputs the differential signal between two inputs: the image signal which the sensor detects when the LED does not emit light (SIG1) and that which the sensor detects when the LED emits light (SIG2). As a result, the background image is canceled and only the target image is obtained. SIG2 has the image of the target illuminated by the LED in addition to SIGl which includes only the background image. The 3D image of the target can be obtained simultaneously by measuring the intensity of the reflected light from the target, because the intensity of the reflected light depends strongly on the distance between the target and the LED. Figure 7.7.2 shows the circuit diagram of the CMOS image sensor.The sensor consists of pixels, column differential circuits, row selection shift register (V-Register), column selection shift register (H-Register), a timing generator, and 8b ADC. Each pixel has a photo-diode (PD), a reset transistor (RG), transfer transistor (TG), row selection transistor (ROW) and amplifying transistor (AMP). The pixel has two storage capacitors (Cl,C2) and two sampling transistors (Sl,S2) for storing the two signals described above. Each column differential circuit consists of a slice transistor (SL), a slice capacitor (C4), a differential signal capacitor (C3) and two reset transistors (RG2,RG3).The two pixel storage capacitors not only extract the target from the background object (for example, a wall of a room) but also obtain a n image which is independent of the illumination condition. Moreover, two capacitors in each pixel solve the problem that the extraction is not accurate when target objects move rapidly, because the interval between the two integration periods is much shorter than one frame period. Figure 7.7.3 shows sensor timing diagrams.Here, RG(n) represents RG pulses in n-th line. There are two photo-integration periods and the differential signal readout period in one frame cycle. In the photo-integration periods, all the pixels are operated simultaneously. In the readout period, the differential operation is line-by-line. In the first integration period, only the background si...

People naturally use gesture to aid inter-personal communication. However, during computer interaction, users are obliged to use conventional devices such as keyboards, mice, and game-pads. In this paper a motional interface is described that allows processing of gestural inputs through the medium ofa 3D image-input device, or Motion Processor, which provides a more natural framework for human-computer interaction. Several PC applications for this real-time motionai interface will be presented in the live demonstration session. These applications are edutainment prototypes that make use of natural hand movement.

Our objective was to develop a compact system for training watchmen in their rounds, suitable for use on a standard computer. A virtual space was created by means of computer graphics. The most important aspect of learning the watchman's rounds is familiarization with the route, including distances, directions, and visual features. A method in which visual and aural cues were used to give a sense of traversing a distance equal to that in the real space is described. Tests showed that when the visual information included a simulation of the watchman (the “virtual user”) moving just ahead of the trainee and when a slight rolling of the visual field was used to simulate the motion of walking, the trainees had the sense of having covered a distance close to the actual distance represented in the simulation. © 1997 Scripta Technica, Inc. Electron Comm Jpn Pt 3, 80(6): 74–82, 1997

No abstract