Dragon Systems has released two new versions of their continuous dictation system called NaturallySpeaking, the Preferred Edition and the Deluxe edition. Both are priced somewhat higher than the original personal edition, which is still available.
Most significantly, they now include the ability to record the speaker's voice behind the recognized words so that one may play back what one has dictated. Moreover, when one selects or clicks on a word to correct it, that word or phrase can be played back and be heard.
The new editions also have speech synthesis capabilities allowing any material which appears in text on the screen to be read in a computer voice. They both allow multiple users to have each with their own voice file.
The Deluxe edition allows the development of topical vocabularies for a particular subject area or a particular speaker which should contribute to greater recognition accuracy within a constricted domain.
Microphones
Always, one is reminded that the position of the microphone is crucial: Being very near the tip of the mouth, very close but slightly off to the side to avoid some of the breathing noises, the interpretation of which could result in some unwanted extra words.
In general, Dragon has made a strong and largely successful effort to recognize extraneous sounds including those created by the speaker, as blank output. This is a highly useful feature.
Microphone technology continues to improve for use with speech recognition. Emkay, a division of Knowles Electronics, have designed very attractive wireless head set microphones, with two-way communication, which should be available soon.
These wireless headsets, in addition to being very light, do not require the wearing of any additional transmitter pack on the belt or elsewhere. Everything is contained in the headset which is nevertheless very light and a long-lasting battery for all day use.
We tested the Emkay radio headset which appeared to have a very clear tone with good noise cancellation. Emkay also puts out a full line of conventional wired headsets which are reported to have excellent noise cancellation characteristics.
VXI continues to offer a full line of microphone headsets for speech recognition use, including a telephone/computer switch or device. Dragon often ships VXI microphones with their product.
Andrea continues to offer a full line of microphones, and is planning to release a radio headset microphone for speech recognition and telephony.
Shure also offers a full line of microphones, as does Sennheiser.
Software
A number of products have been released which enhance the usability of IBM and Dragon Systems, the two continuous general English speech recognition systems available at this time.
Typhoon Software offers a product using IBM ViaVoice which integrates continuous speech recognition with WordPerfect, including the use of command and control operations.
Corel/WordPerfect recently announced an agreement with Dragon in which Dragon Systems will be integrated into their suite of software applications, and should be available soon.
SpeechTrieve offers a simple development program called SpeechLinks for integrating speech recognition into any Windows 95 or NT product. This allows a person to speech enable a software program running in NT or Windows 95, adding dictation capability as well as command and control by voice.
VAST (Voice Activated Systems Technologies), a medical dictation product utilizing Dragon Systems, offers customized dictation of notes and records for psychiatry and some other medical specialties, including built-in databases of official diagnoses, medications etc.
Philips speech recognition partners also offer specialized medical and legal dictation systems. VAST and Philips speech recognition systems can be networked in server-client versions.
We are expecting the imminent release of Lernout & Hauspie's new continuous general English speech recognition system pioneered by Kurzweil, and supported now by investment from Microsoft.
Bill Gates has said publicly that Microsoft may be able to incorporate speech recognition into some future editions of their operating system to allow more complete integration.
Stay tuned here for further announcements.
Peter Fleming, a consultant in speech recognition, can be reached at aris@world.std.com (617)923-9356.