Speech Technology Magazine

 

More Powerful Chips Allow Better Systems at Lower Prices

One of the most critical components in sparking the further development of speech recognition technology is the dramatic growth in the speed and power of microprocessor chips.
By Lizanne Kaiser - Posted Jan 31, 1997
Page1 of 1
Bookmark and Share

One of the most critical components in sparking the further development of speech recognition technology is the dramatic growth in the speed and power of microprocessor chips.

Each generation of chips has marked the commercialization of a more advanced group of speech recognition systems and technology. Philips Dictation Systems was able to develop the first commercial continuous speech dictation system largely as a result of the increased speed of the Intel’s 80486 processor, and the Pentium and Power PC chips have allowed for the development of intelligent recognition systems and continuous speech with large vocabularies.

Most recently, it appears at press time that Motorola is prepared to release its much heralded 820 and 821 chips designed specifically for speech applications. One company insider reported that the new development from Motorola represents a big step forward in performance as the Motorola chip can, running at 25 megahertz, do work previously requiring 75 or more.

The increase in microprocessor power has also led to dramatic price declines, not just in speech recognition, but in all facets of the computer hardware industry. Storage, memory, and peripherals have all been impacted by the changes.

Because of the much lower price, speech recognition is now possible in many consumer products where it would not have been achieved less than five years ago.

Voice Direct from Sensory Inc.
Sensory Inc. has introduced the Voice Direct speaker-dependent, speech recognition chip that recognizes spoken words with greater than 99% accuracy.

Voice Direct allows low-cost consumer electronic products, such as telephones, personal digital assistants, lights, and remote controls, to incorporate high quality speech recognition quickly, easily, and cheaply.

Speakers train the chip to recognize specific words or phrases by repeating each word twice. Voice Direct will recognize and respond to sets of one to 60 words. Voice Direct includes optional English prompts that provide users with helpful spoken instructions. International language speech phrases can also be added.

Keith Kitani, director of marketing for Sensory, said “The Voice Direct chip gives customers another option in creating interactive products that can hear and speak. With the array of products we now offer, customers can choose the right speech technology for their markets, from personal electronics to biometrics security devices.”

Voice Direct is an off-the-shelf device that is ready to use when purchased. No internal programming is necessary. Voice Direct supports two operational modes: stand-alone and external host. In the stand-alone mode, Voice Direct is controlled by instructions from an external microcontroller. Voice Direct is designed to be easily incorporated into existing or new products.

Voice Direct is an ASIC with an on-chip A/D converter and signal processor that accepts speech input for analysis by the chip’s neural-network recognition engine. Speech templates are stored off-chip in SRAM, accessed through a parallel interface, or in EEPROM, accessed through a serial interface. The parallel memory interface is also used to access off-chip international language ROMs. To generate audio outputs, Voice Direct has a speech synthesizer and DTMF generator, both of which use an on-chip pulse-width modulator.

Voice Direct is the fourth member of Sensory’s Interactive Speech line of chips, which include the RSC-164, RSC-164i, and Voice Password. The Interactive Speech chips are designed to provide low cost, high quality speech and audio technology to the consumer electronics market. The RSC chips are general purpose microcontrollers that run advanced speech technologies, including speaker independent speech recognition, speech and music synthesis, and audio/record playback. The Voice Password chip provides speaker verification for security devices.

The packaged chip price for Voice Direct is $5, in quantities of 100,000 units. Engineering samples will be available in the third quarter with production volumes available in the fourth quarter. Prototyping and demonstration kits will also be available in the fourth quarter.

For more information, contact Sensory Inc., 521 East Weddell Drive, Sunnyvale, CA., 94089-2164, or www.SensoryInc.com.

ChipCorder Allows Safe Driving
Information Storage Devices, Inc. (ISD) recently announced that its voice record and playback chips, featuring ISD ChipCorder technology, have been integrated into two new European products: the InfoDrive Systems, a revolutionary new auto safety product manufactured in Italy, and Benefon Sigma NMT 450I a cellular phone from Finland.

The InfoDrive System is a breakthrough new automotive safety product that lets automobiles equipped with the InfoDrive System warn each other of dangerous road conditions. Automobiles with the InfoDrive System will be programmed to send a signal to other InfoDrive-equipped cars if there is an accident, roadway obstruction, or even a rapidly decelerating car within 500 meters. The InfoDrive System will transmit a signal within a two kilometer radius.

Upon receiving the signal, the InfoDrive System will use ISD’s 20-second record and playback chip (ISD1420) to play a pre-recorded message, warning other drivers of the danger. Communications between the two systems will not be restricted by language. For example, warnings transmitted by an Italian speaker can be received and played in English.

“The InfoDrive System, with a recorded message enabled by ISD’s ChipCorder technology, is designed to let drivers calmly anticipate risky situations they can’t see due to blind corners or bad weather conditions,” said Maurizio Pugnotti, project manager at AITEC.

The cellular phone application in Finland is the Benefon Sigma NMT 450I analog cellular phone, which uses the ISD 1416, ISD’s 16-second record and playback chip, to enable an answering machine function. This feature lets the cellular phone owner record a greeting in his or her own voice by simply pressing a button. Incoming callers hear the greeting and then dial in their phone number, which is recorded and stored in the phone. To return the call the cell phone owner simply presses a button to activate the autodial feature for automatic callback.

For more information, contact ISD, 2045 Hamilton Ave., San Jose, Calif., 95125, or at http://www.isd.com.

DSP Resource Board from Analogic
Analogic Corporation recently announced the TAP-802, the latest addition to its TAP-800 family of Digital Signal Processing (DSP) Resource Boards, which is designed to meet the major computational requirements of remote data access and Internet telephony.

The TAP-802 open DSP platform supports both the Multi-Vendor Interface Protocol (MVIP) and Signal Computing System Architecture (SCSA) standard interfaces, 128 local RTDM channels, and is available with complete UNIX and Windows NT-compatible software.

Developers can port their applications to the TAP-802 and gain access to the entire computer telephony market. The board is priced at $6,995.

For more information, contact Analogic Corporation/CDA, 8 Centennial Drive, Peabody, MA 01960, or call 508 977-3000, ext. 3521.

OKI Semiconductor Debuts Single-Chip Voice Recognition Processor
OKI Semiconductor has responded to the demands for voice recognition in personal computers, cellular phones, portable equipment, and automotive electronics with the introduction of a single-chip Voice Recognition Processor (VRP).

“OKI’s VRP delivers a ready-to-go voice recognition capability with no need for user training,” said Avi Zakai, product marketing manager. “With its high recognition rate, the VRP is better than current multi-chip solutions, particularly for personal computers, cellular phones, portable equipment, and automotive electronics. Unlike voice recognition software, this VRP doesn’t get in the way of PC or workstation performance,” he added.

OKI’s VRPs also provide user-selectable feedback (either a voice prompt or a simple beep) to verify that a voice request has been understood. Processed with state-of- the-art technology optimized for low-power portable applications, OKI’s VRPs exhibit excellent tolerance to environmental “noise,” which further enhances their high recognition rate.

For more information, contact OKI Semiconductor, 785 Mary Avenue, Sunnyvale, CA 94086-2909, or call 800-OKI-6388.

Page1 of 1