DTMF Detection

Hi all. I�m trying to implement the goertzel algorithm in C#, but without
good results. 

The steps I'm trying are:

1) I perform a goertzel for each of the frequencies in the table (lows and
highs). For Low ones I use N=212 and 106 for High ones.
2) Save the highest ones (Low and High) 
3) Then I perform 4 more goertzel for Low Freq (+/-)1.5%(+/-)2Hz and the
same for High ones, to verify 2nd Harmonics
4) I search in the table the corresponding number and show it

I'm using a sample rate of 8000 and a buffer size of 8K

The program show me numbers without any logic, I mean. When there are no
DTMF, the system, tell me they are.

Which is the minimum value to compare to, I mean. Is there a minimum
threshold to know if the magnitude for this freq are valids?.

Thanks in advance

dkurman wrote:
> Hi all. I�m trying to implement the goertzel algorithm in C#, but without
> good results. 
> 
> The steps I'm trying are:
> 
> 1) I perform a goertzel for each of the frequencies in the table (lows and
> highs). For Low ones I use N=212 and 106 for High ones.
> 2) Save the highest ones (Low and High) 
> 3) Then I perform 4 more goertzel for Low Freq (+/-)1.5%(+/-)2Hz and the
> same for High ones, to verify 2nd Harmonics

If you're going to use the second harmonic test, you only need to test 
two frequencies.  If you get one and only one hit on the row freqs, and 
one and only one hit on the column freqs, then (and only then) should 
you bother with the second harmonic test, and it should be the 2nd 
harmonic of the two freqs that were "hot."

But even with this improvement, there's a better way (see below).

> 4) I search in the table the corresponding number and show it
> 
> I'm using a sample rate of 8000 and a buffer size of 8K
> 
> The program show me numbers without any logic, I mean. When there are no
> DTMF, the system, tell me they are.
> 
> Which is the minimum value to compare to, I mean. Is there a minimum
> threshold to know if the magnitude for this freq are valids?.
> 

Yes.  A better approach than the second harmonic test is a Parseval's 
relation test.  If the two DTMF frequencies contain ~80%-90% of the 
signal's total energy, it's DTMF (assuming twist tests pass).

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
Today is the last day of your life so far.

Thanks for the response. I�m going to try this. By the way, you said if the
two DTMF frequencies contain ~80%-90% of the signal's total energy is DTMF,
but how can I find out this total energy?, I mean, I only have the energy
corresponding to the low freq and High freq, but not the total one.

Thanks again

>dkurman wrote:
>> Hi all. I�m trying to implement the goertzel algorithm in C#, but
without
>> good results. 
>> 
>> The steps I'm trying are:
>> 
>> 1) I perform a goertzel for each of the frequencies in the table (lows
and
>> highs). For Low ones I use N=212 and 106 for High ones.
>> 2) Save the highest ones (Low and High) 
>> 3) Then I perform 4 more goertzel for Low Freq (+/-)1.5%(+/-)2Hz and
the
>> same for High ones, to verify 2nd Harmonics
>
>If you're going to use the second harmonic test, you only need to test 
>two frequencies.  If you get one and only one hit on the row freqs, and 
>one and only one hit on the column freqs, then (and only then) should 
>you bother with the second harmonic test, and it should be the 2nd 
>harmonic of the two freqs that were "hot."
>
>But even with this improvement, there's a better way (see below).
>
>> 4) I search in the table the corresponding number and show it
>> 
>> I'm using a sample rate of 8000 and a buffer size of 8K
>> 
>> The program show me numbers without any logic, I mean. When there are
no
>> DTMF, the system, tell me they are.
>> 
>> Which is the minimum value to compare to, I mean. Is there a minimum
>> threshold to know if the magnitude for this freq are valids?.
>> 
>
>Yes.  A better approach than the second harmonic test is a Parseval's 
>relation test.  If the two DTMF frequencies contain ~80%-90% of the 
>signal's total energy, it's DTMF (assuming twist tests pass).
>
>-- 
>Jim Thomas            Principal Applications Engineer  Bittware, Inc
>jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
>Today is the last day of your life so far.
>

Ok. Now I have changed the way I get the results. First of all, I get the
goertzel of all the freq (lows and highs) and then I find the highest two
magnitude. This two freq, should be one from high and the other from low,
in case we have a DTMF incoming. But here is the problem. When I press 1 I
have these two freq as: 697 and 770 (both from low) and this should be 697
and 1209.

Is this correct or I'm doing this in a wrong way?.

Thanks

>Thanks for the response. I�m going to try this. By the way, you said if
the
>two DTMF frequencies contain ~80%-90% of the signal's total energy is
DTMF,
>but how can I find out this total energy?, I mean, I only have the
energy
>corresponding to the low freq and High freq, but not the total one.
>
>Thanks again
>
>>dkurman wrote:
>>> Hi all. I�m trying to implement the goertzel algorithm in C#, but
>without
>>> good results. 
>>> 
>>> The steps I'm trying are:
>>> 
>>> 1) I perform a goertzel for each of the frequencies in the table
(lows
>and
>>> highs). For Low ones I use N=212 and 106 for High ones.
>>> 2) Save the highest ones (Low and High) 
>>> 3) Then I perform 4 more goertzel for Low Freq (+/-)1.5%(+/-)2Hz and
>the
>>> same for High ones, to verify 2nd Harmonics
>>
>>If you're going to use the second harmonic test, you only need to test 
>>two frequencies.  If you get one and only one hit on the row freqs, and

>>one and only one hit on the column freqs, then (and only then) should 
>>you bother with the second harmonic test, and it should be the 2nd 
>>harmonic of the two freqs that were "hot."
>>
>>But even with this improvement, there's a better way (see below).
>>
>>> 4) I search in the table the corresponding number and show it
>>> 
>>> I'm using a sample rate of 8000 and a buffer size of 8K
>>> 
>>> The program show me numbers without any logic, I mean. When there are
>no
>>> DTMF, the system, tell me they are.
>>> 
>>> Which is the minimum value to compare to, I mean. Is there a minimum
>>> threshold to know if the magnitude for this freq are valids?.
>>> 
>>
>>Yes.  A better approach than the second harmonic test is a Parseval's 
>>relation test.  If the two DTMF frequencies contain ~80%-90% of the 
>>signal's total energy, it's DTMF (assuming twist tests pass).
>>
>>-- 
>>Jim Thomas            Principal Applications Engineer  Bittware, Inc
>>jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
>>Today is the last day of your life so far.
>>
>
>
>

dkurman wrote:
> Thanks for the response. I�m going to try this. By the way, you said if the
> two DTMF frequencies contain ~80%-90% of the signal's total energy is DTMF,
> but how can I find out this total energy?, I mean, I only have the energy
> corresponding to the low freq and High freq, but not the total one.
> 

Sum the squares of the time-domain samples, then take the square root.

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
Dyslexics of the world, UNTIE!

dkurman wrote:
> Ok. Now I have changed the way I get the results. First of all, I get the
> goertzel of all the freq (lows and highs) and then I find the highest two
> magnitude. This two freq, should be one from high and the other from low,
> in case we have a DTMF incoming. But here is the problem. When I press 1 I
> have these two freq as: 697 and 770 (both from low) and this should be 697
> and 1209.
> 
> Is this correct or I'm doing this in a wrong way?.

A DTMF "1" should not produce two low frequencies.  It should instead 
produce the lowest row freq and the lowest column freq (I don't remember 
their values).

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
Dyslexics of the world, UNTIE!

Thanks. I know this, but is what I get from sample. I think the error is in
the method I implement. I choose N = 205 so for this I dispose a sound
buffer of 820 (205*4) at a sample rate of 8000
Then I execute the goertzel for each of the 16 freq.

I didn�t find any step by step source code, and I don�t know where the bug
is.

>dkurman wrote:
>> Ok. Now I have changed the way I get the results. First of all, I get
the
>> goertzel of all the freq (lows and highs) and then I find the highest
two
>> magnitude. This two freq, should be one from high and the other from
low,
>> in case we have a DTMF incoming. But here is the problem. When I press
1 I
>> have these two freq as: 697 and 770 (both from low) and this should be
697
>> and 1209.
>> 
>> Is this correct or I'm doing this in a wrong way?.
>
>A DTMF "1" should not produce two low frequencies.  It should instead 
>produce the lowest row freq and the lowest column freq (I don't remember

>their values).
>
>-- 
>Jim Thomas            Principal Applications Engineer  Bittware, Inc
>jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
>Dyslexics of the world, UNTIE!
>