Convert voip audio to text for debugging

Question

While working on voip apps, I usually end up picking up one phone, talking to it, picking up the other phone and check if I hear myself. This even gets trickier if I'm doing apps with three way calling.

Using a softphone doesn't help.

Ideally, I want to be able to run multiple instances of some command line based sip ua wherein i can dial a number. Once the ua has dialed and the other party ha picked up, both agents exchange audio. But instead of having to hear some audio, the apps instead display some text which identifies the other end. Possibly some frequency pattern that can be converted to text. Then this text is displayed on the app.

Can something like this be done? I'm creating apps against freeswitch. Ideas how to debug voip apps are also welcome in the comments.

Answer 1

yes, absolutely. The easiest would be to have a separate FreeSWITCH server that is used for placing the test calls and sending/receiving your test signals.

tone_stream will generate the tones at frequencies that you need: https://freeswitch.org/confluence/display/FREESWITCH/Tone_stream

tone_detect can detect the frequencies and execute actions, or even better, generate events that you can catch over an ESL socket: https://freeswitch.org/confluence/display/FREESWITCH/mod_dptools%3A+tone_detect

The best way to generate such calls is to use a dialer script which communicates to FreeSWITCH via Event Socket. Here you can see some (working) examples that I made with Perl:

https://github.com/voxserv/rring/blob/master/lib/Rring/Caller/FreeSWITCH.pm -- this is a part of a test suite tat I build for testing a provider's SIP infrastructure. As you can see, it connects to FreeSWITCH, starts event listener, and then originates the call and also expects an inbound call. It then sends and analyzes DTMF.

https://github.com/voxserv/freeswitch-helper-scripts/tree/master/esl -- these are special-purpose dialers, you can also use them as examples.

https://github.com/voxserv/freeswitch-perf-dialer -- this one generates a series of calls, like SIPp does.

Answer 2

Another technique is to play a sample audio file and record the audio being received on the other end[call recording] and then comparing the two. This system works on setup where systems are located at various places and you are testing end to end quality.

There are lot of Audio Comparison tools [like PESQ] should help you not just detect the presence of Audio but also give stats about the degradation of various parameters in the audio stream.

This can be extended to do test analysis of FS patches as and when they are released and also for other hooks or quality standards you want to enforce.