Digital Microphone

[CaiusX]

Hi @nilspupils, @DD4WH and the rest of the gang

I have rigged up two RPi4's recording next to each other, the one my main birds.naturestation.net public system using the Boya-M3 USB mic and the second using the Seeed 2 Microphone HAT system

Location is sub-optimal and I haven't fiddled too much with the ALSA settings for each (same as in #39).

The SEED needs its drivers to be loaded and installed (NOTE USE THE 64bit installer install_arm64.sh).

BirdNET-Pi needs to manually set the card since it isn't recognised out of the box.

In the terminal, get the card number with arecord -l

Card 3 in this case - and put this into the birdnetpi.config -

Restart all services and away you go.

Sooooooo.... what are the results like?

I'll let the reader make their own minds up (note that neither of my installations are hi-fi, all I'm targeting is workable CHEAP systems)-

Both systems detected this BronZe Mannikin -

The two sound files are:

Boya:

_By_Date_2022-04-30_Bronze_Mannikin_Bronze_Mannikin-76-2022-04-30-birdnet-11_05_40.mp4

SEED

_By_Date_2022-04-30_Bronze_Mannikin_Bronze_Mannikin-92-2022-04-30-birdnet-11_05_30.2.mp4

The Boya USB clearly has more "definition" across the whole frequency spectrum, is louder and less noisy. Interestingly the Seeed digital gave a higher confidence for the same sample.

Todays readings for both systems to now are:

(sadly I had the midnight crash on my Boya system so only have recordings from mid - morning. I will post a full days comparison & multiday when available).

Early indications are that both systems have the same top 5 species (if you ignore the oxpecker & bulbul).

Oh AND - the SEED hat has

• a "BUTTON" (imagine what that can be programmed to do - like headless installation and setting in the middle of the Jungle outside of wifi access) and
• additional plugs for GPIO and I2C interfacing, (temperature/humidity/light.......)

It would be great if others could try out the digital solution and help make it better (reduce noise/buzz, find optimal ALSA settings etc). If this is combined with the Analyze algorithm on a RPi Zero W. A self contained standalone off the shelf system for around $30!!

Just imagine.... one could get a 10 unit BirdNET-Pi-Weather Station array for the same price as whatever you want to spend $400 on.

Best
African Penguin

[SlavKarach]

This could be a very stupid question. I do not know much about audio at all. I barely even listen to music to be honest. However, what setup would I need for my RPI to achieve results similar to my iPhone app? Is your setup close to that? My phone detects more species more accurately at any time.

[CaiusX]

Hi Slav

The birdnet phone app doesn't work for me in South Africa, very few local species in the database. The BirdNet-Lite model has over 6,000 global species.

The beauty of BirdNET-Pi is having an easily interrogated database and sound sample history. It's an evolving system & the more users and suggestions/inputs from users and developers, the better it will get.

[SlavKarach]

Hi CaiusX,

I just checked my Merlin app (also made by the Cornell group) and looks like they have a group pack for SA. Did you try that one?

[CaiusX]

Yip, with poor results in the sound recognition section, photo id is good.

[ehpersonal38]

This could be a very stupid question. I do not know much about audio at all. I barely even listen to music to be honest. However, what setup would I need for my RPI to achieve results similar to my iPhone app? Is your setup close to that? My phone detects more species more accurately at any time.

What's your minimum confidence? The app shows results even if there's a very low chance of it being right whereas BirdNET-Pi has a default of minimum 0.70 confidence.

[nilspupils]

Out of curiosity, i ordered the Boya and will give it a good test along with my other digital micrphone candidates soon. It would be great to find a solution which works without an interface card.
(Btw. my guess is that the Boya must have some kind of interface chip in the plug, but i have actually no idea how it works. I will investigate.) Perhaps someone else can answer.

[CaiusX]

Also give the Seed 2 Mic Hat a whirl if you can - quite inexpensive and has a button :)

[djdunc]

Hi @nilspupils am moving over here from this thread in #39 - on a new install RPi 4 (Bullseye Lite 64 bit) I followed the adafruit install guide plus the extra tweaks in alsamixer linked on that page to setup the software volume control - set it to about 58%. Initially it was at 3% so was not recording anything. With this setup I could then record sound, copy it across to a local machine and play it fine. So the mic works. Then installed BirdNet-Pi and based on tips in this thread tried various configurations of the audio card (default, 0, 1 - I noticed the card number sometimes changes on reboot..., even tried the card name) but the audio recordings are coming out silent. It "feels" like the correct audio input is not being set but I cannot figure out why.

[djdunc]

cheers @CaiusX - have stepped through those. I think it is down to some of the recording settings used in the arecord call being made by bird net - tonight I will dive into making some changes and see what I can find.

[mcguirepr89]

Hi, @djdunc

From your output, it looks like you should try these, in this order:

1. dsnoop:CARD=sndrpii2scard
2. usbstream:CARD=sndrpii2scard
3. hw:CARD=sndrpii2scard

Please take another look at my example for why I've chosen those three.

I hope that helps!

[djdunc]

@mcguirepr89 thanks for help on this - finally have working version. To summarise for others (@nilspupils ) steps I had to go through were:

1. install mems mic according to Adafruit example https://learn.adafruit.com/adafruit-i2s-mems-microphone-breakout/raspberry-pi-wiring-test (note I used BirdNet appropriate build for OS Lite 64 bit)
2. add the volume control to my alsa config by editing the .asoundrc file (at bottom of link in step 1) I set my volume in alsamixer to about 60%
3. in the Birdnet tools > settings > advanced settings I set the audio card value = dmic_sv
4. the last step was a hack - I had to edit the BirdNET-Pi/scripts/birdnet_recording.sh file to edit the arecord -f S16_LE setting to be S32_LE - ie:

    if [ -z ${REC_CARD} ];then
      arecord -f S32_LE -c${CHANNELS} -r48000 -t wav --max-file-time ${RECORDING_LENGTH}\
	      --use-strftime ${RECS_DIR}/%B-%Y/%d-%A/%F-birdnet-%H:%M:%S.wav
    else
      arecord -f S32_LE -c${CHANNELS} -r 48000 -t wav --max-file-time ${RECORDING_LENGTH}\
        -D "${REC_CARD}" --use-strftime \
	${RECS_DIR}/%B-%Y/%d-%A/%F-birdnet-%H:%M:%S.wav
    fi

Am now comparing the recordings against an Ultramic 192. The spectrograms look reasonably similar but the detections seem to be fewer - but will leave them running side by side for a day.

So, initial reaction is I will probably not use the MEMS and will look for a USB mic - but I will compare the results from a days listening tomorrow!

MEMS to left, Ultramic to right :

[nilspupils]

@djdunc: Awesome you got it working and thanks for the detailed instructions! The spectrogram looks really good, i wonder how it performs in the long run.

[djdunc]

Final update 24 hours later - both mics placed outside, pointing in more or less the same direction:

I2S MEMS mic:

UltraMic

MEMS was £6, Ultramic £200 - whilst the former was a little more fiddly, it performs pretty well. BTW also tested the AudioMoth in USB mode which worked really well - will be trying to get hold of some of their uMoths.

[srchild]

I can't make comparisons as it is all that I have tried, but as I had it already I tried using a Rode Smartlav+ microphone which I got a while ago on eBay for a good price and seems to work ok. (Smartlav+ has a higher specification than Smartlav). It can take the Rode foam or fluffy wind protection.

It has a TRRS plug so I connected it to ugreen interface like this (though it was cheaper on eBay).

[nilspupils]

@srchild: Thanks for the info and great you got it working with your setup. I think we will make a list of good/working microphones and soundcards to get these informations sorted in one place. Please post this also to here (for analog microphones and soundcard setups):
#39