I'm trying to implement a FSK demodulator in R. The audio is some text data, and as an exercise I converted the text to raw binary (0s and 1s) to be able to check the output of the demodulator. So far I've been able to come up with some demodulated data, but something is amiss: the length of the real data is 6720 characters, but the length of the demodulator output is 1345049, about 200 times longer. So basically there's a whole bunch of extra 0s and 1s (e.g. instead of 0101001 or something, it looks like 00000111110000011111000000011111).
This makes sense because the duration of the audio file is 30.5 seconds, and multiplied by the sampling rate (44100), the result is 1345050.
My question is: what's the proper way to recover data from the demodulated signal? Some process like demod[1,200,400,600,800...]; but then do I start with the first bit or the 200th, or in between? Or have I completely missed the target?