Phraselator is a wireless phrase recognition device developed for Voxtec by integrator, Applied Data Systems (ADS) and is being used to help medical and military personnel in Afghanistan.
Applied Data Systems integrated the Phraselator, with Microsoft's operating system, speech recognition technology from Voxtec, on ADS' RISC-based Bitsy. The Phraselator is a portable and ruggedized PDA-sized device with large-vocabulary, and speaker independent phrase recognition; its first application will be for special military forces in Afghanistan.
Effective deployment of military forces in peacekeeping is often limited by soldiers' ability to communicate. During operation Desert Storm, mass surrenders of enemy troops overtaxed the coalition's ability to provide medical services.
Battlefield doctors improvised a solution using laptop PC's with speech recognition software, and created a list of 'frequently asked questions.' When a phrase was recognized, a WAV file of the Arabic equivalent was played. While vocabulary and portability was limited, the system proved useful.
A team headed by Ace Sarich, a former Navy SEAL, realized that if this function was packaged into a handheld or belt-mounted computer, its utility would dramatically increase. Sarich continued the product development with Voxtec, a developer of voice-to-voice language translation technology. It designs, develops, manufactures and markets a family of hardware and software products. .
Context specific, speaker independent software is commonly used today for 800 directory assistance, flight reservations and other uses. But those applications require server-sized computers.
With the goal of bringing speech recognition into mobile embedded devices, Voxtec contracted with Applied Data Systems to integrate the Phraselator.
This unit had to be ready for 'instant on' phrase translation (eg, Stop or I'll shoot) while preserving full day battery life. Applied Data started with a 32bit RISC CPU capable of performing workstation — like speech analysis on a much reduced power budget.
After a certified process for up-rating for temperature specifications, Intel StrongARM provided the core of this system. Applied Data engineered the other components, including a hi-fidelity, directional stereo audio channel with both inbound and outbound capability to maintain this low power specification.
Windows CE was a natural choice for the operating system offering both multimedia capability and communication support for wireless applications. The system was developed in CE 3.0 to facilitate compatibility with PocketPC 2002 applications, and future version will also be made available with CE.NET to encompass more robust network security and higher performance future CPU's.
A militarized PDA has more requirements than the 'traditional' speech recognition, including the ability to run in the toughest ruggedized environmental conditions, over extended temperature ranges in the harshest heat, rain and snow.
The touchscreen display would have to be readable in full sunlight and full darkness. A power system had to be designed that could run hours off multiple types of batteries (disposable, as well as rechargable). These batteries would have to be able to be charged from various levels of applied current including 12 and 24V DC vehicle power and 110 or 220V AC.
Special care had to be given to the audio system design, allowing full range frequency response from microphone through CODEC and back out to the speaker. A goal for the speech recognition was set at 98% accuracy in near real-time. Finally, the system had to be flexible and built to accept generally available third party hardware and software.
Due to the urgency of demand following September 11, ADS' engineering expertise was critical in making a fast-turn, fully functioning computer in a few weeks, rather than months or years.
The company has specialized in a variety of PDA-type systems for applications ranging from industrial to assistive technology handhelds.
Initially, the Phraselator will be deployed for the military but future applications in public safety and health are anticipated. Police, fire and emergency EMS can use the Phraselators' ability to translate into virtually any language, and the utility that letting people develop large, domain-specific phrase libraries.
The unit is use in Afghanistan speaks contains Dari, Pashto, Urdu and Arabic language modules. The unit is designed to handle 500 to 1,500 phrases in 2 to 10 languages. Eight weeks after development was started prototype units where taken to Afghanistan to develop practical phase libraries and record native languages.
Voxtec is a subsidiary of Marine Acoustics which five years ago began working on a computer-driven language translation device, with funding of nearly $3 million from the Defence Advanced Research Projects Agency (DARPA).
Other target uses for the Phraselator include airport, event and building security where the Phraselator will allow questioning of individuals in any language while maintaining a wireless link to back-office security systems where profiles and questions can be analyzed.