Haven't used this particular library, but have recently completed a hw project using a ST H7 + I2S microphone as a decibel meter and some ML stuffs.
technix outlined the high level steps for you pretty well above. If that doesn't make sense to you, you might do some research or ask some questions until it does. I can comment on the low-level stuff a bit...
So i'm able to receive PDM data from the I2S interface. But it's still unclear to me if the Sound Meter Library can work with PDM or if the signal should be transformed into PCM signal?
As has been said, if you have an I2S microphone and things are working, you better have
I2S PCM data, not PDM. Are you using DMA? Using the HAL? Have you been able to pull some clips and import them into e.g., Audacity (Import --> Raw Data) and they sound sane?
I can think of two things that might be tripping you up. First is that if you have mono data (only one microphone) then half of your samples will be empty because I2S is nominally stereo. You will have to discard the empty channel data unless the library handles this for you (doesn't seem like it). Second, what is the I2S frame format of your microphone and how to do you want to process (e.g., int16, int32)? There is probably a bit of a mismatch. Most I2S mics I've worked with are 24-bit. This means that if you want to process as 4-byte samples you will need to sign extend your data. Unless you are going straight to float, but that wouldn't be compatible with your sound meter library.