A Viable Alternative to Typing

by fowlertm1 min read6th Jun 201328 comments


Personal Blog

I'm thinking about writing a more substantive post about how humans work and how we can work better, a little like this one.  As is common with these sorts of things, once I started to do research and pull on various threads, it turned out that the field was pretty deep and would require time to understand.  But in the meantime, I just thought I would link to this video of someone programming using only their voice. 

As I suffer with symptoms of carpal tunnel syndrome, this is of particular interest to me.  Once I watched it I decided to start looking at different voice recognition software so that I could still get some work done while typing less.  I'm happy to say that even the default software for speech recognition which came with windows is actually very able and accurate.  I dictated almost this entire post using that software.

As far as I can tell, Dragon Naturally Speaking is the gold standard in voice recognition software.  It does come with a pretty hefty price tag, but it may be worth it if you have serious repetitive stress injuries, or as a preventative measure if you're someone who spends a lot of time at their computer.  And if that doesn't work, chances are good your computer has adequate software pre-installed.  

28 comments, sorted by Highlighting new comments since Today at 6:24 AM
New Comment

I was considering setting up a brain-computer interface for typing. There are a lot of existing BCI hardware systems out there, and you could hook them into Dasher), a program for the disabled that makes it easy to point a mouse or finger at the keys they want to type, and makes likelier keys easier to hit.

Idea free for the stealing.

Still, keyboards are pretty effective (I use Pentadactyl to browse the web using just the keyboard), so I would think you could find a better keyboard somehow that could work for you.

There are a lot of existing BCI hardware systems out there

Could you give an example/link? It never occurred to me that this might be something I could buy as of now.

The Emotiv epoc headset is pretty decent. I've worked with it at my university with a couple of BCI set ups, though I've not worked with their software, so I wouldn't know how good that is.

This is a device I got in December 2010, for IIRC $130. It detects tension ("grrr"), eye movement, and some other brain waves. In my experimentation with it, I found the brain wave modes hard to control, while the others were easier.

There was another model I saw that was $300, but I can't remember the name. The amazon page should get you started on finding related hardware.

Note: I haven't actually used a BCI with Dasher, it was just an idea I had for hands-free typing (and possible improvements in costs over existing tools for the disabled).

Both speed and accuracy of brain-computer interface for typing are pretty bad. What your motivation for using it?

I was exploring it

a) as a possible alternative (and very cheap) typing mechanism for the disabled that hadn't been tried before, and

b) because I figured I could improve on it to the point of making brain-typing competitive, or at least make BCI competitive with mouse usage.

Brain-computer interfaces for the disabled have been tried. There's plenty of academic work on the topic. For some people who are completely paralyzed the technology allows them to communicate by typing 1 character per second.

Right, I found that information at the time, but wasn't convinced this was the best achievable performance for such individuals (let alone price-performance), considering what should be possible with consumer-grade BCIs + Dasher.

I still can't convince myself that this is the best they can do. Personal project time?

I think the 1 character per second speed is even done with EEGs that are much better than consumer grade equipment.

It could be possible to do better but it probably won't be easy.

Sure, but I don't think EEG quality (in terms of lab vs. consumer grade) is the real bottleneck; I think it's minimizing the amount of input that must be provided at all by exploiting the regularity of the input that will be provided. The techniques available here may have been overlooked.

I played a bit with Emotiv and find a maximum of one character-per-second pretty believable - at least, if you stick to actual brain signals and not signals from face muscles ( and even with face muscles one character per second seems in the right ballpark).

One character is not the same as one byte of (maximally compressed) information. The whole point of programs like Dasher (and word suggestion features in general) is to take advantage of the low entropy of text data relative to its uncompressed representation. Characteristic screenshot

Were you using a static, non-adaptive, on-screen keyboard? If so, that's why I would think connecting it to Dasher should result in a speed greater than one char per second, at least after the training period (both human training, and character-probability-distribution training).

If someone is looking for something like Pentadactyl but for Chrome, I recommend Vimium. It adds tons of keyboard shortcuts that make keyboard-based browsing awesome.

I highly recommend this ebook and other stuff on trigger points if you suffer from RSI etc. As far as I can tell, trigger points are the biological basis for RSI, and by reading about them and following self-massage instructions, I was able to resurrect my career and my sanity.

As someone who has suffered considerable wrist pain in the past, I would suggest you check out the works of Pete Egoscue, whose exercise methods were a lifesaver for me.

If you are in as bad a shape as I was, you'll likely have to spend many hours a week to recover, but I've found maintenance to be much easier; many months I don't even spend an hour on it, even though I spend even more time on my keyboard than ever. (But the initial recovery process was hellish, I don't mind telling you. Mostly because I hated spending so much time on the exercises, but also because of the existing pain and fear of pain on my part.)

Even so, given the choice between even the best voice recognition and being able to type without pain, I choose the latter.

What sort of hours are you logging on your keyboard these days? It looks like Mr. Egoscue has quite a catalog, could I impose on you to narrow down which books you found helpful, please? I live in South Korea, so getting a non-kindle book shipped over the ocean is a non-trivial undertaking.

What sort of hours are you logging on your keyboard these days?

I spend all day at my computer, but not all of every day is spent typing. One or two days a week I program most of the day, and a few weeks ago I spent part of a week writing about 40 pages of raw text, and then editing and desktop-publishing it into a 50 page ebook.

could I impose on you to narrow down which books you found helpful, please?

I found the Egoscue Method (his 1993 book) to be most informative and motivational regarding the actual functional anatomy. Some of his newer works have shorter sequences of exercises to accomplish similar goals, but the total time difference isn't substantial -- a few minutes shaved off an hour's work. For whatever reason, I never found Pain-Free At Your PC very motivating to me personally to do the routines in it, though I did try some of the exercises.

Excellent, thanks for the recommendation and for your information. I've bought his book "Pain-free at your PC" and I've started doing the exercises. so far the results are encouraging, but it's too early to tell whether or not this will be the solution I've been after.

Have you heard of the dvorak keyboard? There's evidence to suggest it also helps with carpal tunnel.

Sure, but relearning how to type would be a massive investment in time. I want to try as many other solutions as I can before I dive into the Dvorak setup. BTW, can you link to the evidence that it helps with CTS?

Huh, it looks like there aren't any experiments done either way.

If you google "carpal tunnel dvorak," you'll get a ton of people saying it helped, but all the experiments with dvorak so far are done with typing speed, not pain. Of course, that's anecdotal evidence, so take it with a massive grain of salt.

How many wpm are you achieving? How bad it is with rare names and semi made-up words?

It depends, but so far I'm definitely getting fewer words per minute than when I type. As far as strange names or made of words, it tends to do OK but there are times when you have to do a lot of proofreading. It handles technical vocabulary surprisingly well, and since I'm only using the default program that came with my computer, I'm sure that using something like dragon would work even better.

As far as I can tell, Dragon Naturally Speaking is the gold standard in voice recognition software. It does come with a pretty hefty price tag,

A quick google suggests that the basic version costs 99$. Given how much time most people on this website spent typing I don't see how that cost is very significant.

One thing I wonder about voice recognition is the "control characters" issue. If you're typing, you can make capital letters by pressing "shift". A voice recognition system could allow someone to say "shift" to mean that the next word starts with a capital letter, but then what if someone wants the actual word "shift" in their document? There are a few work-arounds that I can think of, but I'm curious as to what approach the programmers went with.

There are cameras that can tell where someone is looking, so it seems to me that one input method would be to have a keyboard where you look at the letters that you want to type.

There are cameras that can tell where someone is looking, so it seems to me that one input method would be to have a keyboard where you look at the letters that you want to type.

This seems way slower than typing or speech recognition.

[-][anonymous]7y 1

In Dragon you can easily enter control characters by saying "Press" or "Type" as in "Press alt", "type control alt". The same seems to be true of most keyboard alternatives out there these days, as is creating macros for frequently used phrases, commands, whatever, as you can see in the video posted by the OP. It took a while to adapt my speech to speech-recognition-friendly pronunciation (not just providing a corpus of my own pronunciation) but now the # of typos ("speechos?") doesn't seem to be outrageous.

In some cases a small set of foot pedals can also serve as keyboard alternatives for frequently-used keys like shift, control, alt.

I use foot pedals + Dragon when my hands aren't up to the job of keyboarding (a week or two every couple of months). I also found Windows SR technology to be adequate for many things but have the luxury of access to Dragon at work. For me speech input is about 85% as fast as typing in sentence/paragraph formats - although there are SR errors I find a second-pass editing process to correct such errors is more effective than the equivalent of backspacing and replacing as I would normally do with keyboarding. My speed drops to more like 40% in programming or other situations where I use a lot of copy-pasting. Could probably improve this if I had to do it all the time.

I have tried eye-tracking input methods, including remote and head-mounted kit just for fun, and it seems just too cumbersome if there are speech or hand-based input alternatives. Although to be fair I haven't tried any of these methods long enough to reach an improvement plateau.

Swyping may help avoid aggrevating your CTS, easy to learn in any layout, and is faster than speech recognition, in my experience, once you account for the inevitable corrections. Neither is any good for programming, unfortunately.