I remember when I was in high school (many moons ago… in the 90s) one of my friends let me borrow a new program that allowed you to “type” on the computer by “speaking.”

Up until that point, the only time I experienced voice commands was when watching Star Trek or some other sci fi show.

The software is actually still around today – Dragon Naturally Speaking – but back then it was a shell of it’s current self.

Back then, you had to spend a good two hours reading text to the software so it could learn your voice and even then it wasn’t all that accurate.

As it turns out, voices are pretty complex things and it requires a lot of processing power to understand them.

Current limits are still obvious when you use something like Visual Voicemail.  I use it with my Google Voice account, and the only ones it can get 90%+ accuracy with are the voicemails done by a computer.  I can usually get the gist of the message left by regular people, but there are plenty of times where it’s complete gibberish.

Even in 2010, voice recognition still works best in a closed enviromnent.  For example, if you call customer service for just about any major company now, you’ll be greated by an automated system.

In their system there are a limited number of commands available to you at any given time… frequently yes, no, help, operator, and sometimes numbers 1-10 and letters A-Z.  It’s a whole lot easier for a computer to figure out what you’re saying when you’ve got 50 or so choices vs the entire English language.

Another example is my new smartphone.  I have an Android powered phone and Google being all about “search” has a pretty good “voice search” program running on the phone.  There are still plenty of times where it has no idea what I’m saying, but if you’re careful to speak slowly, clearly, and in an enviromnent with as little background noise as possible, it’s pretty accurate.

Speech To Text – Transcription Software

Speech to text software has been making pretty big strides recently.  Right now software like Dragon Naturally Speaking can do an excellent job with software based transcription if it only has once person’s voice to deal with.

It’s still not up to the task of handling multiple voices so meetings, phone calls, or anything with a lot of background noise is still out.

However, if it has to be 100% accurate you’re still going to need a human to do the transcribing.  There are plenty of online solutions like Casting Words that can do general transcription economically.

Voice Recognition Software Manufacturers:

• Nuance Sofware – Dragon Naturally Speaking
• 3M Voicescript
• Windows Speech Recognition
• MacSpeech Dictate
• e-Speaking
• Eloquently Stated
• Wizzard Software – Interactive Voice Assistant
• ViaVoice

Companies like Google have also been spending millions on voice recognition research.  If you purchase an Android phone, you can see some of their early results by using the voice commands and voice search on the phone.

Right now, most consider Dragon Naturally Speaking Preferred and Professional to be the best pieces of software in the voice recognition market.