Sound frequency detection in a beehive

Hi all,

My LoRa beehive is finally taking shape, one area I’m having issues with is sound. Due to transmitting sensor data once an hour (Power and LoRa bandwidth considerations) I can’t transmit actual audio for further analysis.

The ESP32 currently wakes up (hourly) to gather sensor data, I was thinking to use something like this FFT Library\example to generate 8 or more “frequency buckets” across the spectrum and transmitting the values in addition to overall volume each hour, what do you think? Has anyone else done something similar?

I noticed @abaretta posted on the old beehive thread about his commercial product that has a device that’s collecting sound data, I wonder if he is doing something similar? His device is coin cell powered so I assume so. I’m not really sure how he’s working out swarming, queenless etc but I plan to just collect data and then cross reference my observations against the data and see if there’s any patterns.

I have an INMP441 sensor currently within the hive, it’s connected via CAT5 cable directly to the ESP32.

1 Like

Hello chinswain,

a while ago, I was involved in the early stage of a project where we wanted to classify and count flying insects by their wingbeats. This was done on the sensor device. Only aggregates were transmitted via TTN to a backend.
Depending on use (lab or outside) and power budget, interesting samples could be stored for later retrieval or transmission by other means. More can be found here:

So my guess is, you are on the right track. I would try to collect sound samples of the different moods a hive can be in and store them on a micro SD card. You can train a model to detect those states in your hive and send alerts and stats via ttn.
I hope this helps.