Synaptics Inc, the neural networking pioneer founded eight years ago by Zilog Corp co-founder Federico Faggin (CI No 761), already has its neural network chips built into cheque verifiers made by Verifone Inc, but the company yesterday announced its first major product, a better mousetrap. Synaptics reckons that its TouchPad offers a more accurate and reliable way to control personal computers and other electronic appliances than a mouse or a trackball. The TouchPad enables users to control the cursor more precisely and to translate pressure information, replacing the motion of clicking a mouse button, and Synaptics sees it being used in video cassette and cable television remote control units as well as computers. The TouchPad uses neural-based mixed-signal technology and is claimed to converts finger motions into cursor motions in a natural, intuitive way, as well as offering a variety of tapping gestures for button-free clicks and drags. It uses low-cost capacitive sensing technology, with no moving parts, and not only senses top-to-bottom, right-to-left motion, but pressure or area of contact as well: in graphics tablet mode, it becomes a finger-operated pressure-sensitive graphics tablet for a variety of design applications. Finger motion is detected by measuring the finger’s effect on an array of capacitive lines integrated into the TouchPad module. One side of the module is the sensor surface, with electronic components mounted on the reverse. The TouchPad OEM module for systems integration into notebook and computer keyboards is available now at under $20 in large quantities, and Twinhead Corp and Seiko-Epson Corp plan to integrate it into future systems. Synaptics engineers also say they are about a year away from completing their own handwriting recognition system, which they predict will have accuracy above 99%, and is also applying neural network and other adaptive circuits to vision and speech recognition, exploring speech recognition by developing a silicon model of the cochlea spiral-shaped cavity in the inner ear. The company has spent $14m so far, $7m was raised from investors and the rest from contract work for other companies. It holds over 40 patents in VLSI neural networks, analogue circuit techniques, imagers and recognition.
