I think you'll find that this is a phun physics problem. Once you get the raw PCM data, you'll want to run it through a Fourier Transform. I think AGraham has a Fourier library. I suspect that you'll find that blowing on the microphone will generate a very broad, nearly flat energy spectrum. You can characterize the flatness of the spectrum by calculating the standard deviation. I'm not sure if there are statistical libraries available but that's a very easy calculation to do on your own. Then, you can set a range of acceptable standard deviation values that your app will consider indicative of blowing on the mic.
The advantage of this method is that it is volume-independent. You could blow softly or loudly and this method should be able to detect it.