History / Mission
Voice Recognition Systems was founded in 1994 to provide speech recognition solutions to enterprises that utilize transcription departments for documentation and record keeping. Speech Recognition is one of the biggest emerging technologies today for physicians and support staff to properly document patient encounters and notes. The majority of our business comes from healthcare. Those who contact us have either heard about this technology being utilized or sometimes are so far behind in their documentation process they are under pressure to stay caught up or get caught up.
Voice Recognition Systems caters to the individual that wants to produce large amounts of text and documentation in the minimal amount of time. What we have to provide however is the proper "mix" of software and hardware to accomplish that task. This takes a thorough knowledge of computer operating systems and it takes time if you want to get the best results. Giving a client the proper solution and providing them with consultation and understanding is what we do best!
Advancements in Technology
What we have seen during the last 19 years are leaps and bounds in speech recognition technology especially in speed and accuracy. The first speech recognition system commercially available was called DragonDictate and it was the system we started with in 1994. It was based on what was called discrete speech recognition and you had to speak slowly and clearly one.....word.....at.....a.....time! It ran on a 386 processor with 16 MB (that's megabytes not gigabytes) of RAM, cost $5,000.00, ran on the Microsoft DOS operating system and did not include the DragonDictate software which then sold for $995.00 and was classified as a discrete speech recognition system with a 30,000 word general purpose vocabulary and required 20-30 minutes of initial enrollment training in order to use it. We didn't have the caliber of computer processors and memory that we have today so we had to squeeze all the power and performance out of a machine that we could. RAM/Memory was extremely expensive then at $50.00 per megabyte and our first notebook we demonstrated DragonDictate on had $800.00 in memory alone! Click here and you can see a demonstration of what is was like in January 1995.
The first continuous speech recognition product was Dragon NaturallySpeaking 1.0 introduced in 1998 and we are now on NaturallySpeaking Version 12 as of September 2012. What we've seen is a gradual increase in recognition accuracy with the release of every subsequent version of Dragon. In the earlier versions of Dragon you spent more time "training" your voice to the system and the recognition accuracy got better over time. Now we are getting 99% accuracy with about four minutes of initial voice file training! It's not your grandfather's Buick anymore so if you've used Dragon in the past and it didn't work so well you'll see it's a much different product now than it was even five years ago! Click here to see our current demonstration of Dragon NaturallySpeaking 12 Professional.
The two basic rules of speech recognition are:
- The faster the processor, the faster you can dictate
- The more memory (RAM) that you have, The more accurate your dictation will be!
Recommended Requirements (a.k.a - What you need to do if you want it to work the way it's supposed to)
For a client we recommend a system with no less than a Core2Duo processor, 4 gigabytes of RAM and a good noise cancelling microphone. As long as we have these 3 things present, we can be reasonably sure that your recognition accuracy is going to be good. However there is a difference in getting "good" speech recognition accuracy and "excellent" speech recognition accuracy! Several years ago we published a document called 12 reasons Voice Recognition Systems users are successful and we now have an updated version of that document available here.
What we've learned over the last 19 years is that proper hardware optimization is as equally important as owning the equipment itself. If software is not installed properly, the machine you are running your applications on is not optimized properly and/or you don't have the right training toolbox and system utilities, speech recognition accuracy will suffer. Every case we have ever encountered where a problem arises with speech recognition accuracy boils down to a few things which include incorrectly configured, poorly optimized or under powered machines, improper installation, improper training, low quality microphone (almost always the case with off-the-shelf software) or in more rare cases the end user was not motivated enough to use speech recognition to achieve a positive outcome.
A different approach:
We decided to take a different approach to the speech recognition market. When we first looked at this technology 20 years ago we saw that it had promise because if you think about it you have 2 different forms of input into a computer and that's a mouse and keyboard. Speech recognition is just a third form of input into the computer and can be used interchangeably with the mouse and keyboard. If you watched the demonstration from January 1995 above you'll notice that it worked decently even then. An average American English speaker engaged in a friendly conversation speaks at a rate of approximately 110–150 wpm. We know that most people don't type that fast and it made sense to us that anybody who needed to generate a lot of documentation within a limited time frame could make good use of this technology.
20 years ago however this was an expensive technology because computers were expensive and software was expensive. Speech recognition presented a unique problem because you needed a lot of processing power and memory to run it with decent accuracy. Dragon has over 1,000,000 lines of code and every time you speak an utterance you are running a very complicated set of algorithms to determine a best match of what was said out of a possible 160,000 word database. A lot of people say the same things in different accents and dialects and this is why you initially train a 4-5 minute enrollment file so that a speech model (file) can be built on your particular accent or dialect.
What we found out early on was that if we optimized and tweaked the computer hardware that Dragon was running on we were getting much better accuracy (often 30% or better) then other users were getting if it wasn't done. In different tests that have been conducted over the years an average person who types at 50 WPM can produce a three-page, 900 word document in approximately 18 minutes but a person dictating at 140-160 WPM can produce the same 900 word document in approximately 6 minutes. The question was however how many corrections were you going to have to make using a speech recognition program to have a final, finished document?
It became apparent to us that although we knew the technology was there the recognition accuracy wasn't. What we realized was that getting good speech recognition accuracy was not going to rely on software alone but rather a combination of the right computer hardware, software, training, system utilities and a toolbox to educate users on how to use the application properly. We decided the best way to approach this was to offer a turnkey solution which included everything. This way we can control the quality of the build, parts, installation and training and our clients are very successful using speech recognition technology because of this approach.
To our knowledge Voice Recognition Systems is still the only company that provides complete speech recognition turnkey solutions that include everything ready to go out of the box. You can do a google search on speech and/or voice recognition turnkey solutions and you'll see we come up 5 or more times on the first page. Although our turnkey systems are more expensive then you will find in the retail channel these are systems that are built with the best quality parts, professional level software, professional noise canceling microphones, best system utility software and 12-15 hours is spent getting everything installed and optimized perfectly before delivery to the client. Every solution is a custom solution tailored to the needs of the client. What has to be taken into consideration is not the initial investment but the time and money saved in document generation because of better accuracy and turnaround time. This is now more important than ever within healthcare because providers are now doing a lot more work for less reimbursement.
Voice Recognition Systems has been successful with thousands of installations over the last 19 years. Our approach is different because we have the knowledge and industry partners behind us to know how to put all the components and knowledge together for a successful end-user experience. Call us at (800) 631-8255 or (859) 266-2800 and we'll be happy to discuss a solution to your dictation or transcription needs.
us for more information