Welcome to the Forum for InventorSpot.com, the most popular invention related website in the world. Read our welcome message.
by krugalug » Thu Jul 19, 2007 1:05 am
- krugalug
- White Belt
- Posts: 1
- Joined: Thu Jul 19, 2007 12:11 am
Chris,
I just saw you on American Inventor tonight. Your idea looks great, and I know that you will pursue it with great enthusiasm. I'm also sure that after everyone sees you on TV you will be getting calls from many many of them, and I hope that some of them are people who can help you bring your invention to life.
I'm not an inventor, and I'm not even very smart. I am an electrical and computer engineer working for a computer company in Austin, and my expertise is in computer and server design (only 7 years). Anyway, I wanted to say a few things about your idea, and you can take them for what they're worth.
I know that I don't have to tell you this, but forget about what they said on American Inventor about your idea not having mass appeal. You know as well as anyone that your invention would be a huge breakthrough for millions of people around the world. Add a little more software to this product, and you could turn this into a translator and talk to people who don't even speak English. I'm sure others can think of even more uses for it, but I don't want to focus on trying to improve the great idea that you already have.
You can break your invention down into four parts:
On the main user's end:
1) An interface for a user to enter language into the device without using speech. There are thousands of mobile devices that already do this. Most use either a numeric phone keypad text messaging interface (slow and clumsy if you ask me), a written character recognition interface (Palm Pilot uses a pad and a stylus pen), and the most useful is the keypad that you will find on a Blackberry or Sidekick. I think this technology exists, so it should be easy to implement into your invention.
2) A display to the user. This is a simple screen that displays the text of the language that the "other" user has entered on their side. A text screen like this is also easy to implement in any portable device.
On the other end:
3) An output for the "other" user. This will most likely be a speaker, but it could just as easily be a text screen if the other user is also deaf or maybe just happens to be in a situation where it is difficult to hear; a loud bar, for example. Again, this is fairly simple technology to implement.
4) This is the difficult one. It is the speech input that the "other" user uses to enter language into the device. On American Inventor you said that this technology exists. I think you are right, to a certain extent. I have read a few magazine articles about this throughout the last 10 years, and I even remember some of my better professors discussing the subject of speech recognition back when I went to school in '99.
Speech recognition is much more difficult than it seems, otherwise I think we would see more of the technology all over the place. You can take the example of calling up the airlines to find out when your flight leaves, and they ask you to say your flight number and answer some YES or NO questions. They may ask you to repeat yourself, or they sometimes get the numbers wrong, but it works pretty well overall. Remember that it is a very limited vocabulary that it must decipher, so being able to recognize regular speech is probably much more difficult.
All of that being said, I do think that it is VERY possible to take the best speech recognition technology today and implement it into your invention. The biggest hurdle might be that many companies have put a lot of money into creating this software, so they may want liscensing fees, but it should be possible. It may also be a little more limited than what we hope for today. The software may not recognize heavy accents well or may need to be trained to only understand voices that it has already heard. And people may need to speak more slowly than they do in regular conversation. Even with those limitations, I think this product would be great, and the voice recognition software will only get better as time goes by, so it will eventually be as perfect as you hope.
I've rambled quite a bit, but I'd like to say that you can probably begin some limited experiments using a regular computer. You can enter text in and have the computer "say" it through the speakers. I think this software may already come bundled in some of the Microsoft Windows packages, and I wouldn't doubt that Apple also has this. As for the speech recognition, I think your best bet is to scour the internet for all speech recognition software out there and try it out. You should go for the best (and probably most expensive) software out there. You may be able to call the companies and use your American Inventor fame to get it for free. Once you find something that seems to work using a regular computer, the next challenge is shrinking it all down into a portable device. That discussion is for another time, but I know that if you have some sort of working model using a laptop computer, you will have a better chance of convincing the big companies to manufacture a portable device that does the same thing. It's all about the proof of concept.
I just saw you on American Inventor tonight. Your idea looks great, and I know that you will pursue it with great enthusiasm. I'm also sure that after everyone sees you on TV you will be getting calls from many many of them, and I hope that some of them are people who can help you bring your invention to life.
I'm not an inventor, and I'm not even very smart. I am an electrical and computer engineer working for a computer company in Austin, and my expertise is in computer and server design (only 7 years). Anyway, I wanted to say a few things about your idea, and you can take them for what they're worth.
I know that I don't have to tell you this, but forget about what they said on American Inventor about your idea not having mass appeal. You know as well as anyone that your invention would be a huge breakthrough for millions of people around the world. Add a little more software to this product, and you could turn this into a translator and talk to people who don't even speak English. I'm sure others can think of even more uses for it, but I don't want to focus on trying to improve the great idea that you already have.
You can break your invention down into four parts:
On the main user's end:
1) An interface for a user to enter language into the device without using speech. There are thousands of mobile devices that already do this. Most use either a numeric phone keypad text messaging interface (slow and clumsy if you ask me), a written character recognition interface (Palm Pilot uses a pad and a stylus pen), and the most useful is the keypad that you will find on a Blackberry or Sidekick. I think this technology exists, so it should be easy to implement into your invention.
2) A display to the user. This is a simple screen that displays the text of the language that the "other" user has entered on their side. A text screen like this is also easy to implement in any portable device.
On the other end:
3) An output for the "other" user. This will most likely be a speaker, but it could just as easily be a text screen if the other user is also deaf or maybe just happens to be in a situation where it is difficult to hear; a loud bar, for example. Again, this is fairly simple technology to implement.
4) This is the difficult one. It is the speech input that the "other" user uses to enter language into the device. On American Inventor you said that this technology exists. I think you are right, to a certain extent. I have read a few magazine articles about this throughout the last 10 years, and I even remember some of my better professors discussing the subject of speech recognition back when I went to school in '99.
Speech recognition is much more difficult than it seems, otherwise I think we would see more of the technology all over the place. You can take the example of calling up the airlines to find out when your flight leaves, and they ask you to say your flight number and answer some YES or NO questions. They may ask you to repeat yourself, or they sometimes get the numbers wrong, but it works pretty well overall. Remember that it is a very limited vocabulary that it must decipher, so being able to recognize regular speech is probably much more difficult.
All of that being said, I do think that it is VERY possible to take the best speech recognition technology today and implement it into your invention. The biggest hurdle might be that many companies have put a lot of money into creating this software, so they may want liscensing fees, but it should be possible. It may also be a little more limited than what we hope for today. The software may not recognize heavy accents well or may need to be trained to only understand voices that it has already heard. And people may need to speak more slowly than they do in regular conversation. Even with those limitations, I think this product would be great, and the voice recognition software will only get better as time goes by, so it will eventually be as perfect as you hope.
I've rambled quite a bit, but I'd like to say that you can probably begin some limited experiments using a regular computer. You can enter text in and have the computer "say" it through the speakers. I think this software may already come bundled in some of the Microsoft Windows packages, and I wouldn't doubt that Apple also has this. As for the speech recognition, I think your best bet is to scour the internet for all speech recognition software out there and try it out. You should go for the best (and probably most expensive) software out there. You may be able to call the companies and use your American Inventor fame to get it for free. Once you find something that seems to work using a regular computer, the next challenge is shrinking it all down into a portable device. That discussion is for another time, but I know that if you have some sort of working model using a laptop computer, you will have a better chance of convincing the big companies to manufacture a portable device that does the same thing. It's all about the proof of concept.
by 5rocks » Thu Jul 19, 2007 5:22 am
-
5rocks - Green Belt
- Posts: 213
- Joined: Wed Dec 27, 2006 12:36 pm
- Location: milwaukee
Great job Chris K. It was nice to see someone who is deaf make it on American Inventor. I too am partially deaf and wear hearing aids. I thank God for the hearing that I do have and cannot even begin to know what it is like to not have any hearing. People like you give others hope and that is what it is all about. When I was on the show last year, they interviewed me about my hearing loss and I think they wanted me to get emotional, but I stuck to my goal of just showing them my product. Keep up the good work. Good things are coming your way!
Pat
www.thetakeaseat.com
Pat
www.thetakeaseat.com
by Road Show » Thu Jul 19, 2007 9:59 am
-
Road Show - Brown Belt
- Posts: 962
- Joined: Mon Nov 20, 2006 5:39 pm
- Location: SoCal
Krugalug, you say you aren't that smart, but you are well above average to be in the field you're in. Anyway, I think you are right in terms of your assessment of Chris's invention; that the voice recognition technology will be the biggest challenge. Clearly, the current uses for this technology are in situations where certain responses to questions are anticipated by the device. The wider the scope of anticipated responses, or the within the scope of completely unanticipated conversation, the less effective the technology becomes. Sure, there is probably the perfect speech recognition software somewhere out there, but I would think, like you, that it might be cost prohibitive at this time. A few years from now, though, it might be as common as solar powered calculators.
RSG
RSG
by AmericanCynic » Thu Jul 19, 2007 5:07 pm
- AmericanCynic
- Yellow Belt
- Posts: 99
- Joined: Sun Jun 10, 2007 11:07 pm
Don't forget that voice recognition with any degree of accuracy requires a very high signal to noise ratio. That's why all the available software like IBM's ViaVoice and Dragon NaturallySpeaking come packaged with headsets that keep high quality noise-cancelling microphone booms a small and consistent distance from your mouth. With the Voice Inside, you'd basically have to stick the device right in front of somebody's face and tell them to speak and don't move their heads, hoping all the while that the background noise won't overwhelm the software. Not even the best noise-cancelling superdirectional microphones can pick up clear enough sound from a distance. And there will still be errors for the foreseeable future, since context-sensitive voice recognition is very difficult, like the examples of whether homophones should be transcribed as "I scream" or "ice cream," "I speak" or "ice peak" or "ice pick," "arson" or "our son." That may be Chris' biggest weakness in envisioning this device. Since he doesn't know what speech sounds like, he has no idea that many words and phrases sound extremely similar.
Not to mention almost all voice recognition software needs to be trained to recognize the characteristics of a limited number of voices. Even the ones that claim to not need training are imprecise until enough time and corrections have passed that it has learned your idiosyncrasies. Speaker-independent continuous voice recognition remains a pipe dream. Have a stranger sit down at your computer and microphone and you'll likely get gobbledygook out of your voice recognition software. Even if all of this worked, what happens when there are multiple speakers in the room? Unless the microphone is pointed at the right person, Chris will still be the last person in the room to laugh at a joke. There are a great many stumbling blocks to this invention and Pat Croce was right. $50k wouldn't have been enough for a working prototype. Even the million dollar prize wouldn't have been enough. Remember, the big corporations have been in this field for decades and even they can't get flawless speech-to-text on fast desktop computers with lots of storage. Shrinking it down to a handheld with decent all-day battery life would be more difficult than anyone can imagine. I'm not trying to stomp on anyone's dream, but just be sure you're going into this with both eyes open. 2001 has come and gone and HAL 9000 is as distant as he was in 1969.
Not to mention almost all voice recognition software needs to be trained to recognize the characteristics of a limited number of voices. Even the ones that claim to not need training are imprecise until enough time and corrections have passed that it has learned your idiosyncrasies. Speaker-independent continuous voice recognition remains a pipe dream. Have a stranger sit down at your computer and microphone and you'll likely get gobbledygook out of your voice recognition software. Even if all of this worked, what happens when there are multiple speakers in the room? Unless the microphone is pointed at the right person, Chris will still be the last person in the room to laugh at a joke. There are a great many stumbling blocks to this invention and Pat Croce was right. $50k wouldn't have been enough for a working prototype. Even the million dollar prize wouldn't have been enough. Remember, the big corporations have been in this field for decades and even they can't get flawless speech-to-text on fast desktop computers with lots of storage. Shrinking it down to a handheld with decent all-day battery life would be more difficult than anyone can imagine. I'm not trying to stomp on anyone's dream, but just be sure you're going into this with both eyes open. 2001 has come and gone and HAL 9000 is as distant as he was in 1969.
24 posts
• Page 3 of 3 • 1, 2, 3