I'm not the person you asked, but I can tell you that I sing very poorly, yet have gotten SoundHound to recognize songs.

I think its about doing the notes correctly, not the voice. I think it could even work with your mouth closed.

It's all about getting the direction change of subsequent notes right. Up or Down is all that matters. And the nice bit is that it doesn't matter if you transpose or if you go up too far or down too far for the next note.

You'll note that you need to define a sequence of similar notes as 'down' or 'up' to make it work.

I mean, that's what humming is.