For the project page click here.

DTMF stands for Dual Tone Multi Frequency. This is an in-band telecommunication signalling system using voice-frequency band over telephone lines between telephone equipment and other communications devices and switching centres. DTMF is used to represent up to 16 keys (most telephones only use 12 of these). Each key is represented by two different frequencies. The first bin (lower frequencies) consist of frequencies under 1kHz and the second bin (Upper bin) consists of frequencies above 1.2kHz. The combination of the two tones will be distinctive and different from tones of other keys and these tones cannot be mimicked by voice or random signals.

The intent of this project is to design a DTMF Decoder and create a Java API for a it. I started this project while I was on a short internship at VASTech during the December 2015-January 2016 UCT vacation break. My mentor for this project was Albert Visagie (@avisagie).

The API is designed for use in programs where a DTMF signal needs to be decoded (given it is in a valid form of .mp3 file, .wav file or as an array of sample points; either as double[] (mono) or as double[2][] (stereo)).

• DTMF Decoder for .wav and .mp3 files or when given an array of sample points. (double[] / double[2][]).
• The API can decode only mono and stereo channeled audio signals. It separately decodes and returns the DTMF tones found in each channel.
• Has an audio file interface which can be implemented for more audio file types (ogg, wma, etc...)
• DTMF Tone/Sequence Generator that can export to .wav files.
• Goertzel Class which can be used independently with arrays of sample points representing a signal.
• The API includes a GUI Application (Java Swing) which can decode DTMF .mp3 and .wav files and also generate DTMF tone sequences.

• Optimising the Goertzel Class to improve on speed and performance.
• Coming up with a more efficient way to detect noise and human speech to improve rejection and minimise false hits when decoding random noise files.
• Decoder could give a precise location (time) of detected tones within the audio file.
• Implement signal processing techniques that improve detection like correlation to boost the SNR, window functions to reduce spectral leakage, etc... (I hadn't studied these at the time of this project)

Check out this small CMD program that uses the decoder. To use this decoder in your code, import com.tino1b2be.dtmfdecoder.DTMFUtil;

If you have a signal you want to decoded that is saved as a .mp3 or .wav , it can be decoded this way:

DTMFUtil dtmf = new DTMFUtil(filename);
dtmf.decode();
String left_channel = dtmf.getDecoded()[0];
String right_channel = dtmf.getDecoded()[1]; // only works if it exists else it throws an indexing error

Where filename is the path to the .mp3 or .wav file. More file types can be implemented using the AudioFile interface but only these two are implemented so far.

This is particularly useful if you are decoding an audio (or any signal) stream. If you have an array of samples of the signal ( from -1 to 1 with a mean of 0) and where Fs is the sampling frequency of the signal, the decoder can be used this way:

int Fs = 8000;
double[] samples = {/* array of samples */}
DTMFUtil dtmf = new DTMFUtil(samples, Fs);
dtmf.decode();
String sequence = dtmf.getDecoded()[0];

int Fs = 8000;
double[][] samples = {{/* array of samples from first channel */},{/* array of samples from second channel */}}
DTMFUtil dtmf = new DTMFUtil(samples, Fs);
dtmf.decode();
String[] sequence = dtmf.getDecoded();
String first_channel = sequence[0];
String second_channel = sequence[1];

A PDF version of the full report on this project can be viewed here. This report covers everything from the research made in the project, the pseudo code and algorithms used along with the motivations for using them, testing and much more. You can contact me for more information on my email ([email protected]). To find out more about me please visit my website.

The project is licensed under the MIT License.