http://blog.8pixel.net/ An RSS feed for Simpleblog EN
This is my first Blog posting, and since several of my ventures are in the field of speech recognition (also known in the industry as voice recognition), I decided it would be a good first topic to write about. The thing that bothers me most about speech recognition is the resistance people have to adopting this great technology.  Through my seven years of experience in the field I’ve learned that many, if not most computer users are resistant to adopting speech recognition, and there are three main barriers behind this: The time it takes to setup the system, the learning curve, and the need to modify work habits.  If you are one of those people who is overwhelmed by the thought of using speech recognition or if you're considering it, then hopefully you will find my thoughts on the subject helpful.  So here we go:

I’ve found that many trial users who download a speech recognition program don’t want to invest the time to setup and train the speech recognition engine and configure the microphone.  It seems like a daunting task.  Many don’t even bother to read the instructions and don’t know they have to train the speech engine, or they just skip the step.  And some users don’t bother with getting a good quality headset microphone.  A decent microphone only costs about $20.00.  When a user skips training, or uses a poor quality microphone or a desktop or built-in microphone, the recognition accuracy results are usually poor and the user dismisses the product as another download that doesn't work.  The truth is that training only takes about 20 to 30 minutes, and then its done for good.  This is a small investment of considering the long-term rewards.   When the speech engine is trained properly (at least two or three training sessions), a good quality headset microphone is used to insure consistent position of the microphone, and the microphone is configured before each use; the accuracy of speech recognition is typically 95% or better.  Isn’t this a worthwhile investment to enhance your computing experience for your lifetime?

Learning how to use the program and learning the speech commands is another impediment to adopting speech recognition.  To many users, this also appears as a daunting task, but the truth is its easy.  That is if you are using the right program.   A good speech recognition program is built from the ground up with this in mind.  The user interface is designed to be intuitive and developer incorporates documentation that makes it easy for the user to learn the program and speech commands.  The speech programs offered by CoolSoft and SpeechVibe (discussed below) have demo videos (a picture is worth 1,000 words), quick reference guides (a summary of basic speech commands) and detailed instructions (for further reference).  “What can I say?” dialogs, tool tips and “say what you see” features are designed into both of these programs.  The goal is to make it easy for the user to succeed at using speech recognition.  Try it and you’ll see.

The third issue is user related.  To use speech recognition effectively, you have to change your habits.  Most people correct their mistakes as they type.  With speech recognition it is more efficient to dictate your text first, and then make your corrections.  Many users (including myself at one time) are resistant to this change of habit.   But once you get the hang of it, you might find that it truly is much better than typing.  I was at a speech recognition conference several years ago, and one of the speakers presented some statistic.  He said their research showed that the average person types at 35 words per minute with a 5% error rate.  The average user of speech recognition achieves a rate of 50 words per minute with the same 5% error rate.  This represents a whopping one-third increase in productivity.  I’d like to see any other type of program match that. 

Two more notes about speech recognition accuracy.  First, a speech engine cannot make spelling mistakes.  The recognition errors come from recognizing one word for another (like “pest” instead of “test”).  This is great for people (like me) who aren’t good spellers.  Second, a speech program that has a correction feature and utilizes the user adaptation capabilities of a speech engine will learn when a recognition error is corrected.  Then, the speech engine is unlikely to repeat that mistake in the future.  SpeechVibe is one such program that makes use of this feature.

For those of you interested in trying speech recognition, I recommend SpeakToText, available at http://www.CoolSoftLLC.com as a good program to introduce yourself to the world of speech recognition.  It’s a simple, easy to use program that satisfies the needs of most users, providing much of the functionality of high priced speech recognition software at a fraction of the cost. 

Once you’re hooked, you’ll want to upgrade to SpeechVibe or SpeechVibe professional.  I’m working on a plan where users of SpeakToText will get a special upgrade price.  SpeechVibe is a really kick-ass speech product that in my opinion gives Dragon Naturally Speaking a good run for its money.  Try it, and you’ll see what I mean.  Personally, I think it far outperforms the Windows Vista native speech recognition (yes, Windows Vista has speech recognition built-in).  The best part is that SpeechVibe works in Windows 2000 and Windows XP – so you don’t have to, umm, upgrade, to Vista to get speech recognition.  For example, one of my friends installed Windows XP on his new Vista machine and called that an upgrade.  You can get SpeechVibe on the CoolSoft website or go directly to http://www.SpeechVibe.com Checkout the demo videos, download the program and try it. 

Looking forward to your comments.  Until next time, all the best,

Paul]]> 9/25/2008 3:16:00 PM http://blog.8pixel.net/?view=plink&id=116 116