The ALIVE space utilizes a visual recognition system called Pfinder, short for person finder, developed at the Media Lab for tracking a person's hands, face or any other color-discriminable feature [4]. Pfinder uses an intensity-normalized color representation of each pixel in the camera image and multi-way Gaussian classifiers to decide which of several classes each pixel belongs to. Examples of classes are background, left hand, right hand and head. The background class can be made up of arbitrary scenery as long as the color value of each pixel does not vary beyond the decision boundary for inclusion in another class. The mean of each cluster gives the coordinates of the class, and the eigenvalues give the orientation. Pfinder provides updates on each class roughly 6 times a second. Further details on the visual recognition system can be found in [4].
The information from the visual recognition system is used to steer a fixed beamforming algorithm. Azimuth calculations are performed from the 3-space coordinate data provided by the mean of the head class. The re-calculation of weights for each new azimuth is a relatively low-cost operation since the weight update rate is 5 Hz.\
The use of visual recognition techniques makes it possible to achieve both the optimal signal-enhancement performance of a fixed beamformer with narrow beam width and to get the spatial flexibility of an adaptive beamformer.