Waking ESP32 using sound

Thread Starter

Brettjohnson7191

Joined Apr 28, 2022
27
Hello all, me and my group are working on a senior design project and for part of our project we are trying to make a glass break sensor that has far less false alarms. To do this we have plans to use an ESP32 and a microphone to detect the sound of the glass breaking. It will then go through a pattern recognition process to compare the audio file to known glass break signals stored on a micro SD. However still in the design phase this process seems to use a lot of power.
We had thoughts of placing the MC into a sleep mode and waking it using an interrupt if a suspect sound file is sent. We are wondering if anyone has any ideas how this could be done. We have suggested using a second lower power MC for the initial screening of the sound and the esp32 for the final decision.
 

MrSalts

Joined Apr 2, 2020
2,767
Hello all, me and my group are working on a senior design project and for part of our project we are trying to make a glass break sensor that has far less false alarms. To do this we have plans to use an ESP32 and a microphone to detect the sound of the glass breaking. It will then go through a pattern recognition process to compare the audio file to known glass break signals stored on a micro SD. However still in the design phase this process seems to use a lot of power.
We had thoughts of placing the MC into a sleep mode and waking it using an interrupt if a suspect sound file is sent. We are wondering if anyone has any ideas how this could be done. We have suggested using a second lower power MC for the initial screening of the sound and the esp32 for the final decision.
The ESp32 can achieve low power but the ESP32 modules contain more than just the microcontroller. Depending which module you're using, the voltage regulator and memory both consume power. On top of that, the ESP32 isn't the fastest starting microcontroller - especially if you initiate the wifi connection in the setup. This slow startup means you may miss the glass break sound.
 

Ian0

Joined Aug 7, 2020
9,670
The “attack” (as in attack decay sustain release) is probably one of the most identifiable parts of the sound. If you miss it then it will make the identification so much more difficult.
I‘d have a micro running all the time to process the audio and just use the ESP32 for comms.
As there is a lot of high frequency components in the sound of breaking glass, you‘ll need a high sampling rate. That micro is going to be busy, and its power consumption won’t be small.
 

MrSalts

Joined Apr 2, 2020
2,767
I think it would be much easier to use a low power microphone with a band pass filter and tune it for a particular frequency and intensity and simply have that output to wake the microcontroller via the enable pin.
 
Top