zimolzak / wav-in-python

Learning and messing with audio import/processing in Python

Python Frequency Shift Keying Analysis

Decodes a WAV file of frequency shift keying (FSK) noises to its bitstream.

Bare-bones but cross-platform alternative to Sorcerer decoding software. fldigi software does FSK but not enough raw bitstream inspection for me.

Can be used for STANAG-4481 "NATO" radio signals. Specifically, STANAG-4481F. The ones I received are at 50 baud, with a frequency shift keying (FSK) or "RTTY" shift of 850Hz. Detection of the mark and space tones (and the shift) is automatic. Allows reshaping bitstream into various size rectangles, to allow inspection for various serial asynchronous framings (like 5N1 or 8N1). Should work on any binary (2-tone) FSK signal (not MFSK), but you may have to change passband filter parameters if your mark/space tones don't match STANAG-4481. Baud must always be set manually.

Command line

python main.py sample-data.wav
# You will see stdout and a png will be created

python main.py signal.wav  # 10 second file
# png overwritten

Within Python

from wave_helpers import whole_pipeline

bitstream = whole_pipeline(infile='sample-data.wav', outfile=None)

# Now do something cool with bitstream. 

Outline of approach

1. Read samples from the WAV file (22,050 values / sec).
2. Apply a short-time Fourier transform to find the most intense tone at any time point (300 values / sec).
3. Reduce the vector of tones to a vector of bits (50 values / sec).

Requirements

• wave (Python Standard Library)
• SciPy
• Matplotlib
• NumPy