I have written software to detect the transmission of frequency sequences. I'm using Geortzel's Algorithm to filter desired frequencies from a data array. The data array is generated from audio capture. I am transmitting a sequence of bits encoded in two frequencies, F-mark and F-space.

Currently, my signal-to-noise ratio is not satisfactory and even with scaling and thresholding of the signals, I am getting too many false positives.

I've had the idea to use two independent frequencies to represent each the mark and the space (4 frequencies in total) and only register a valid signal if both are detected simultaneously. This seems like it would only approximately double the computation time which is acceptable.

Would this improve improve the accuracy of my detection, or would it have no effect on my signal-to-noise ratio? Would it also increase the likelihood of false negatives and missed detections?

  • $\begingroup$ Are f-mark and f-space chosen so that they are orthogonal over the length of the Goertzel filter? Are you doing any synchronization of the Goertzel filter window? $\endgroup$ – hotpaw2 Nov 14 '15 at 19:49
  • $\begingroup$ @hotpaw2 Apologies, my knowledge of signal processing is fairly limited. Could you explain what you mean by orthogonality over the length of the filter? $\endgroup$ – user18279 Nov 18 '15 at 11:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy