I am developing an alerting device for hearing-disabled people. I have a mild hearing loss myself, and I cannot hear

• the front doorbell if I'm in the basement
• the teakettle whistling unless I'm in the kitchen
• the clothes dryer bell unless I'm within 2 meters of it
• the oven buzzer unless I'm within 6 meters of it.

   
The concept is a small battery-powered device to be mounted at a sound source and trained to detect its sound (doorbell, teakettle, etc.) When the sound is detected, an SMS will be sent to the user (e.g. "Doorbell", "Teakettle", etc.)

Using one device near the source for every sound limits noise and simplifies the User Interface. The user just clicks "Listen for" and plays the reference sound. The device then starts compares incoming sounds to the reference sound. If the two are sufficiently similar, it sends the SMS.

I did a software demo at Akron Makerspace on my 10 year old MBP. I can't find a way to fool it. Putting the mic right down the throat of the source means the inverse square law eliminates spurious signals, and the frequency detection just ices the cake!

Other alerting devices typically use one expensive processor per multi-source space instead of a cheap processor at every source. That's why they have to be wired in, or plugged in, or have a single-purpose notifier the user has to lug around, or are a chore to program and adjust -- because the device isn't right at the source.


Ron Phillips

https://www.linkedin.com/in/rwphillips

This is a really interesting idea. I'm thinking you could make a Bluetooth Low Energy or WiFi node network, have the network report to a central processor or computer that could send the SMS. Set each of the nodes to listen for a different sound. Could allow for better portability and flexibility for a user. Love the Akron creativity here.

Could BLE range be an issue? I like the network idea, because apparently WiFi really pulls a lot of power, so a central unit that could be plugged in might save the nodes' battery life.

It certainly could, but if the nodes are strategically placed and can handle master/slave mode simultaneously, I imagine it could be possible bounce a signal from node to node until it hits a node within range of a device that's currently scanning for advertising peripherals. (Let me know if that's worded confusingly.) It's sort of like a mesh network, replacing WiFi with Bluetooth connectivity. As long as one node is in range of another, they can talk to each other and find someone who can talk to you.

I was thinking of developing on the ESP8266, since it has such low current requirements, especially with the WiFi shut off. The WiFi could be shut off unless the device was sending an alert.

There's no need for the device to listen or download from WiFi unless the user is setting the reference sound. Since they have to be at the device to set the reference sound, I was thinking of leaving the incoming WiFi off unless the user pressed a "Setup" button on the device.

Now, if Expressif delivers that new module with both WiFi and BLE . . .

Or maybe the new Pi Zero W, since it has WiFi and Bluetooth built right in. I'll have to look at the power consumption, and how to "sleep" the wireless.

Connecting Smart Hearing Aids to the Internet via IFTTT has an interesting take on the problem. They're thinking of providing IFTTT prompts aurally through Bluetooth-enabled hearing aids via the smartphone linked to the aids.

That would let the user set the phone down within BLE range and get auditory notifications without touching the phone. I think.