Here's my problem. I'm a quadriplegic confined to a power wheelchair, & I control it using a sip & puff system. When I activate the chair, the drive mode latches in the on mode and to stop, I give it the opposite command. The tube that I use to suck or blow air into sometimes will fall out of my mouth and the wheelchair will continue to go until I hit something that stops me. I have broken my left leg twice. I have no control to stop the wheelchair. There is a manual switch that most people can use to stop, but it does not work for me. So, the solution, I believe, would be a voice activated momentary switch with a microphone on my control tube where I could yell something to close the switch then have it open so I could continue when someone gives me the tube again. I'm looking at a sound activated switch but do not think it will work because when I'm in a noises area the chair would keep stopping?? Any ideas?

 

Well to accomplish something like that is going to require a lot of coding of a micro computer, and a bit of know how of audio engineering with circuits.

Personally if I was ahead of a project like that I would first start off by testing out filters to remove everything, such as background noise, popping, and etc. Then when you get a very clear output of the voice. The next step would be to do something like ADC sampling with a microcontroller.

What you can tell what the microcontroller to do is start reading the ADC input pin if it's a certain value. (When we start talking) is to start sampling the input values of the voltage and listen for a word like "computer". When it hears that word it will start sampling the next word you say and figure it out.

The main problem with stuff like this is it could totally backfire at anytime. So it could one day end up being a very seriously unsafe device. Example: Loud background noises, other people talking. It could miss interpolate and could drive you in the middle of the road, but if your in quite environments all the time, I dont see why not. One idea is to add a button you could press so it knows exactly when to start sampling.

Then the next step after you talk it will tell you if it recognized your command. Then you could just press the button as a yes or no.

I personally would jump into a project like this head first if I could, but with no funding and having to focus on two full time jobs, a task like this would be very difficult to accomplish in time before the next thing comes in the way and does it right away.

 

I could add a few things...using 2 microphones wired in opposition to each other will cancel "far field" noises. That means any sound that hits both microphones equally. It's an old recording trick. After that, you have to decide if you can shout louder than any near field noise that might happen and maybe set the "off" point solely on loudness. If that won't work, you're into the ajm113 situation of using an enabling word followed by recognition signal and a command. Then there is the idea that you can't sip or puff a circuit board. Somebody is going to have to build this for you, and it will take days, even if nothing goes wrong.

 

Just brainstorming. When you are use the sip/blow tube to control the wheel chair, do you normally keep the tube between you lips all the time while you are moving? If so, I'm envisioning a proximity/touch sensor in/on the tube to sense that it's in you mouth. If not the chair stops. There are small micro-controllers available now with "touch" inputs.

Ken