Gesture Recognition Technology

Hand Posture and Gesture Recognition Technology
This section discusses the requirements for hand posture and gesture recognition. It describes the two main solutions for collecting the required data to performrecognition, the glove-based solution and the camera- or vision-based solution, and looks at the advantages and disadvantages of each.
Data Collection for Hand Postures and Gestures
The first step in using hand posture and gestures in computer applications is gathering raw data. This raw data is then analyzed by using various recognition algorithms (see
Section 3) to extract meaning or context from the data in order to perform tasks in the application. Raw data is collected in three ways. The first is to use input devices worn by the user. This setup usually consists of one or two instrumented gloves that measure the various joint angles of the hand and a six degree of freedom(6DOF) tracking device that gathers hand position and orientation data. The second way to collect raw hand data is to use a computer-vision-based approach by which one or more cameras collect images of the user’s hands. The cameras grab an arbitrary number of images per second and send them to image processing routines to perform posture and gesture recognition as well as 3D triangulation to find the hands’ position in space. The third way to collect raw hand data is to combine the previous two methods in a hybrid approach with the hope of achieving a more accurate level of recognition by using the two data streams to reduce each other’s error. Very little work has been done on hybrid tracking for hand posture and gesture recognition, but this type of tracking has been successful in augmented reality systems like Auer[7] and State[98], and could well be...

... middle of paper ...

...tizes the voltage output of each sensor and then modifies the value using a linear calibration function.
This function uses gain and offset values to represent the slope and y-intercept of the linear equation. This equation allows software calibration of the glove and thus makes it more robust for a variety of hand sizes. The author’s personal experience and an evaluation by Kessler et al. [50] suggest the CyberGlove is accurate to within one degree of flexion. It works well for both simple and complex posture and gesture recognition (Wexelblat[116] and Fels[35] verify this claim). The only negative in regard to the CyberGlove is its price; the 18-sensor model is available for $9800 and the 22-sensor model for $14,500. But even though the glove is expensive, it is the best available glove-based technology for accurate and robust hand posture and gesture recognition.