How can I detect the number pressed from a dtmf tone (.mat) file?

I have a DTMF tone that is a .mat file, it just consists of one button press on a telephone and i want to be able to detect the number pressed and output it to the command line. How exactly can I do this?

I posted an example of the .mat file

[DGM]

Have you checked any of the suggested tools or answers in the sidebar? --------------------->>

[Elliot Wyllie]

I have, none of them really worked for me and none used a .mat file

[DGM]

A mat file is just a container for any sort of variable that matlab can hold. The mat file just contains a vector and a scalar specifying the sampling frequency.

Using this FEX submission I just randomly picked:

https://www.mathworks.com/matlabcentral/fileexchange/10440-dtmf-encoder-decoder

load('dial_0.mat')
thiskey = decode(y',1,fs,numel(y)/fs)

gives

thiskey =
0

Of course, the documentation and parameterization for this function is pretty terrible, but that's some kind of tradition on the File Exchange.

Answer 1

[DGM]

Basing off of this example:

https://www.mathworks.com/help/signal/ug/dft-estimation-with-the-goertzel-algorithm.html

% this loads two variables:
% fs -- sampling rate
% y -- signal vector
load('dial_0.mat');

Nt = 205;
source_tones = [697 770 852 941 1209 1336 1477]';
symbols = {'1','2','3';'4','5','6';'7','8','9';'*','0','#'};

k = round(source_tones/fs*Nt);  % Indices of the DFT
estim_f = round(k*fs/Nt);      % Frequencies at which the DFT is estimated
tone = y(1:205);

% Estimate DFT using Goertzel
ydft = goertzel(tone,k+1);

% find the two most powerful frequencies
[~,sourcetoneidx] = sort(abs(ydft),'descend');
sourcetoneidx = sourcetoneidx(1:2);

% sourcetoneidx should contain one index from
% each group of source tones (first four, last three)
row = min(sourcetoneidx);
col = max(sourcetoneidx)-4;

% look up the symbol in the table
thissymbol = symbols{row,col}

gives

thissymbol =
    '0'

[Elliot Wyllie]

Interesting, thank you so much for the example. Would there be any way to use FFT as a way to decode DMFT instead of the Goertzel algorithm?

[DGM]

Yes. The example I linked to on the File Exchange uses FFT. Just download it and open decode.m. It's a bit of a mess, but it has comments explaining the steps.

EDIT: it really is kind of a mess. I stripped it down and added some comments so it makes more sense:

load('dial_0.mat');
	
d = numel(y)/fs;

% fft analyze of the frequency coded signalvector
signal=abs(fft(y));                        
% Analyze to the half sampling rate, because the frequencies after are
% not interesting for analyze. They were created by sampling the signal 
x=1:(fs/2)*d;

% plot the spectrum to see the peaks
plot(signal(x))

% maximum of the array
[~,idx]=max(signal(x));
% frequency 1 
f1=(idx/d)-1;

% we no longer want to consider frequencies close to this peak
% set adjacent samples to 0 
idx=idx-10:idx+10;
signal(idx)=0; 

% next maximum of the array
[~,idx]=max(signal(x));
% frequency 2 
f2=(idx/d)-1;

% sort hi/low frequencies
if f1>1000
    var=f1;
    f1=f2;
    f2=var;
end

% include frequency tolerance and sort the key frequencies (the actuall  
% tolerance are at with time: >= 0.0499 s and sampling rate: 3300Hz)
if ((f1>692) && (f1<720))
    row=1;
elseif ((f1>765) && (f1<800))
    row=2;
elseif ((f1>847) && (f1<880))
    row=3;
elseif ((f1>936) && (f1<960))
    row=4;
end

if ((f2>1204) && (f2<1240))
    col=1;
elseif ((f2>1331) && (f2<1360))
    col=2;
elseif ((f2>1472) && (f2<1500))
    col=3;
elseif ((f2>1628) && (f2<1660))
    col=4;
end

% look up the symbol associated with this DTMF pair
symbols = {'1','2','3','A';'4','5','6','B';'7','8','9','C';'*','0','#','D'};
thiskey = symbols{row,col}

The peak-finding and binning routine can probably be cleaned up significantly, but the basic idea is there. Do FFT, find the peaks, do symbol lookup.

EDIT:

Well, I did say it could be cleaned up.

load('dial_0.mat');

% do fft
signal=abs(fft(y));   
d = numel(y)/fs;

sourcetones = [697 770 852 941 1209 1336 1477 1633];

tonewindows=zeros(size(signal));
tonepower = zeros(8,1);
for f=1:numel(sourcetones)
    tonepower(f) = sum(signal(round(d*(sourcetones(f)+(-20:20)))));
    tonewindows(round(d*(sourcetones(f)+(-20:20))))=1000;
end

% plot the signal and sampling windows
x=round((500:2000)*d);
plot(signal(x)); hold on;
plot(tonewindows(x))

% same thing as in the first example
[~,sourcetoneidx] = sort(tonepower,'descend');
sourcetoneidx = sourcetoneidx(1:2);

row = min(sourcetoneidx);
col = max(sourcetoneidx)-4;

% look up the symbol associated with this DTMF pair
symbols = {'1','2','3','A';'4','5','6','B';'7','8','9','C';'*','0','#','D'};
thiskey = symbols{row,col}

This approach is more like the DFT example. Instead of trying to sort out peaks, just look where you expect peaks to be and find the two most powerful ones.

[Elliot Wyllie]

Jeez that stuff goes right over my head, im a bit of a beginner when it comes to coding and signal processing. Im assuming I would have to rework all that code to fit the .mat file because decode file seems to work based off the encode file.

[DGM]

At this point, I wouldn't worry too much about using the FEX code directly. Like I said, it's a mess and requires the user input a bunch of extraneous parameters because it was probably designed only to do a particular thing (yes, probably it was just made to recycle the tones generated by the encoder).

That's why I cleaned it up and reduced it to a simplified script that takes the mat-file as given. No parameters needed, just a filename (assuming your mat-files all contain variables called fs and y).

EDIT:

I cleaned up my suggested code and made a encoder/decoder set of tools. Hopefully these will be a better resource to the next person that has the same task:

https://www.mathworks.com/matlabcentral/fileexchange/91500-simple-dtmf-encoder-decoder

There's no special toolboxes needed, and it should be compatible back to at least R2009b.