Decode DTMF (touch tone) signals against background noise

[RobC]

Hi,

I have written a VI to detect the touch tone '1' DTMF signal (697Hz and 1209Hz) in an input received through a microphone. The VI reads the sound, assembles a waveform and then passes it through parallel bandpass filters tuned to 697Hz +/- 10% and 1209Hz +/- 1% respectively (the tighter tuning on the higher filter came after some trial and error at minimising nuisance triggering). The filtered signals are then squared and the DC signal compared against a constant in order to return a boolean output of whether the tone has been heard or not.

I have just about got this to work so that the tone can be detected without my own speech triggering it. However if I play music in the background this is causing nuisance triggers, presumably as the music contains the two desired frequencies. Also, when I clear my throat loudly that triggers it as well.

Can anybody think of a suitable method I could use to cause a trigger on the 697/1209Hz signal without background noises triggering it? I would like to keep the method as simple as possible as I would eventually like to replicate my VI in low-cost electronics, hence my use of bandpass filters.

Thanks. 

[altenbach]

Maybe instead of comparing the level with a constant, compare it with a level based on the background noise (e.g. obtained from the unfiltered signal or from a filtered signal with a slightly shifted frequency). Also check if the two signals are similar in amplitude, for example.

[PhillipBrooks]

I recalled that there is a minimum signal duration for a DTMF signal. A quick google search on "dtmf singal duration" returned several links stating that the ITU minimum duration is 40 ms.

A simple timer may be sufficient;if two of the eight frequencies are simultaneously and continueously present for > 40 ms, then you have a valid entry...

Message Edited by Phillip Brooks on 03-08-2010 07:40 AM