Dmitry Kiryashov wrote:
>
> Hello Scott.
>
> > Often times the source of the signal provides the noise too. For
> > example, in the DTMF decoder application, I found that the algorithm
> > would produce a fairly tightly banded signal for the 'ideal case'.
> > However, if there is a voice present during the DTMF signal, the decoder
> > would produce a signal that flutuated. Yet the median/average value (or
> > more precisely the average of the median values) over several samples
> > was fairly close to the unperturbed case. In general, anytime there is a
> > sporadic high frequency component present in your signal it's possible
> > that MOMA (Morgan's filter) will simply not respond to it. (Which, as I
> > said before, may be a good thing - depending on the application.)
>
> To detect DTMF in addition to filtering the main frequencies it is
> requeried
> to know that high harmonics of mains are absent (to exclude false voice
> recog-
> nition as DTMF). The simple idea was visit me. Maybe we should add
> another one
> simple wideband filter (for instance starting from 3nd harmonic minimal
> freq
> and ending at 5rd harmonic of maximal freq). If this additional filter
> don't
> detect anything that all is ok. Probably I couldn't offer the
> programming
> details of such suggestion but idea looks reasonable. (Probably it'll be
> some-
> thing related to autocorrelation signal with itself)
>
> WBR Dmitry.

This is the approach that Analog Devices takes in their DSP-based DTMF
decoder. (Actually they decode the 'fundamentals' and the '2nd
Harmonics'.) However, this approach isn't entirely succesful (from what
I've read in other literature). But even if you do implement this
filtering strategy, you're still going to be plagued by the 'broad band'
nature of some conversations. In other words, normal conversation can
overlap the DTMF frequencies.  (The 'autocorrelation' trick helps,
though.)

Scott