Your idea to just play musical notes will likely sound like a kid playing the piano one note at a time. Only one "voice" or "channel" at a time is possible. Any serious or decent musical sound track is going to be a combination of several tracks so as to bring a harmony to the music. The content of the typical WAV file is going to be a mix (summation) of several sources and the stored 16-bits/sample LPCM @44.1KHz is a recording of that mix.

In the very simplest case of a WAV file that was a simple single channel of recorded notes from a plonk - plonk - plonk note sequence from a keyboard it may be possible to decode the LPCM and emit the equivalent notes. For a more complex summation that may even have additional "effects" merged in it will be next to impossible to decode and restore it back to the original note sheet.

If you have no choice but to flog these WAV files then personally I would suggest that you seriously look at just finding a way to have your MCU system just "play" the WAV content. Pick out the most basic WAV format and just decode the file format directly to samples that you feed to a D/A. In the basic LPCM format you'll have about 22usec to handle each sample so expect to need an MCU with something more than a 1 usec instruction cycle rate.

The reason you find a MIDI to notes converter is because MIDI is basically an interface medium that passes notes around. If the simple plonk plonk song is actually suitable to your application then you would then be far better off to demand being supplied a set of MIDI source material. Have the client hire someone with a MIDI studio to source the materials in that format!!

I suspect the the client will not be satisfied with a plonk plonk type of solution. The reason to believe this is because they gave you WAV/MP3 content because of its richer full bodied sound. I'm sure that they would be less than happy if you ended up producing some type of gross plonk plonk emulation of their source material.

From an MCU perspective it will be very unattractive to try to manage any type of WAV format data of any significant duration of time. The storage becomes cumbersome and expensive. For this reason the much more space efficient storage of a MIDI type data stream is very nice for MCU applications.

So in the end I would strongly suggest you stop trying to break your back over converting the WAV/MP3 to notes. Instead insist that the client get busy and provide MIDI data. When they work with someone in a studio they can get a feel for what the thing will sound like before you try to reproduce it. As a result the expectations that they come to you with will be level set to what you can reasonably provide.

Now as a developer it can be very helpful to your client if you provide them with alternatives to consider. One is to show the alternative with the MCU driving one of the various add-on audio chips. Another solution is to offer up a solution with an ARM based SOC that has an audio subsystem built onboard.

Michael Karas