symbol extraction

symbol_extraction.py

symbols_real = []
symbols_imag = []

base_freq = 100                     # This is in Hertz
symbol_rate = 23.6
num_tones = 16                      # This particular MFSK modulation contains 16 tones i.e. (500-100/23.6) =~ 16.95  
                                    # But its apparently odd to have an odd number of frequencies
                                    # So, I settled on '16'. Turns out that was right. 
tone_zero = int(round(base_freq/symbol_rate))       # the first or lowest tone in the sequence 

sample_rate = 48000
nyq_rate = sample_rate / 2.0
order = 9                          # This number is after experimentation. like Filter-design always is
cutoff = 560                       # same as above.

for index in iter(peaks):
    despread_re = np.multiply(us1, re[ index[0]  : index[0] + len(us1) ])
    despread_im = np.multiply(us1, im[ index[0]  : index[0] + len(us1) ])

    normal_cutoff = cutoff / nyq_rate
    # Get the filter coefficients 
    b, a = butter(order, normal_cutoff, btype='low', analog=False)
    real = filtfilt(b, a, despread_re)
    
    normal_cutoff = cutoff / nyq_rate
    # Get the filter coefficients 
    b, a = butter(order, normal_cutoff, btype='low', analog=False)
    imag = filtfilt(b, a, despread_im)

    
    fft_instant_real = np.fft.rfft(real)
    real_fft = fft_instant_real[tone_zero : tone_zero + num_tones]
        
    fft_instant_imag = np.fft.rfft(imag)
    imag_fft = fft_instant_imag[tone_zero : tone_zero + num_tones]
       

    currsymbol = np.argmax(np.absolute(real_fft))
    symbols_real.append(currsymbol)
    
    currsymbol = np.argmax(np.absolute(imag_fft))
    symbols_imag.append(currsymbol)