Hello, I would like to ask more experienced persons for help with the problem I'm working at... I have two audio signals: one outgoing voip stream, and one incoming (from the other side) I would like to compare the streams, especially with regards to: - transport problems (RTT, jitter, packet loss) - acoustic distortion (transcoding between different codecs) First I need to find what is the mean delay between the signals, and trim them to matching starting/ending time. Then I'd like to compare them, but not only to get resulting number but the curve showing the distortion in time I guess the best would be to use DFT to get frequency domain representation and then compare the samples using correlation function. Before I start working on methodics and the programming itself (I have to admit that I only have remains of school knowledge in this area), I would like to ask if there are not any standard procedures, algorithms or software to start with.. Final aim should be to automate this testing using some script and not to click it by hand, but I'm interested also in directions to use some graphical audio processing software I prefer solution on linux platform, but it isn't prerequisite, Scilab/ Matlab is also OK. Any ideas on where to start? with best regards
Comparison two audio signals
Started by ●April 8, 2009
Reply by ●April 8, 20092009-04-08
"kopi" <lboxkopi@gmail.com> wrote in message news:534b4115-c023-46d4-a25a-c58f5159398d@o11g2000yql.googlegroups.com...> Hello, > > I would like to ask more experienced persons for help with the problem > I'm working at... > > I have two audio signals: one outgoing voip stream, and one incoming > (from the other side) > I would like to compare the streams, especially with regards to: > - transport problems (RTT, jitter, packet loss) > - acoustic distortion (transcoding between different codecs) > > First I need to find what is the mean delay between the signals, and > trim them to matching starting/ending time. > Then I'd like to compare them, but not only to get resulting number > but the curve showing the distortion in time > > I guess the best would be to use DFT to get frequency domain > representation > and then compare the samples using correlation function. > > Before I start working on methodics and the programming itself (I have > to admit that I only have remains of school knowledge in this area), I > would like to ask if there are not any standard procedures, algorithms > or software to start with.. > Final aim should be to automate this testing using some script and not > to click it by hand, but I'm interested also in directions to use some > graphical audio processing software > I prefer solution on linux platform, but it isn't prerequisite, Scilab/ > Matlab is also OK. > > Any ideas on where to start? > > with best regardsActually there are quite a few open libraries for programming VoIP applications. Have a look at EMIPLIB: http://research.edm.uhasselt.be/emiplib/emiplib.html It makes use of JRTPLIB which is a library for the Real-time Transport Protocol and RTCP. RTCP will send out sender & receiver reports, and these will give you the transport information you need.
Reply by ●April 8, 20092009-04-08
thanks for your answer I didn't now about JRTPLIB, thanks for the tip, I'll test it for dialing and receiving test calls we are using sripts writen in Python with PJSIP library bindings audio data for comparision we gets directly from RTP connection betwen PJSIP and PBX our measurment enviroment looks like this : src script --- RTP --- src PBX --- VoIP / POTS / GSM / ... --- dst PBX --- RTP --- dst script (src and dst VoIP PBX can be same one) we can get statistics from RTCP, but they are only for RTP connection to PBX, not for whole path our main focus is to measure what happened to audio data between src a dst PBX (recoding between different codecs, conversion to/from ISDN, ...) an that is a reason why we are focussed on audio signal comparison and not on RTP statistics On 8 Dub, 15:18, "Neill" <nore...@mail.com> wrote:> "kopi" <lboxk...@gmail.com> wrote in message > > news:534b4115-c023-46d4-a25a-c58f5159398d@o11g2000yql.googlegroups.com... > > > > > > > Hello, > > > I would like to ask more experienced persons for help with the problem > > I'm working at... > > > I have two audio signals: one outgoing voip stream, and one incoming > > (from the other side) > > I would like to compare the streams, especially with regards to: > > - transport problems (RTT, jitter, packet loss) > > - acoustic distortion (transcoding between different codecs) > > > First I need to find what is the mean delay between the signals, and > > trim them to matching starting/ending time. > > Then I'd like to compare them, but not only to get resulting number > > but the curve showing the distortion in time > > > I guess the best would be to use DFT to get frequency domain > > representation > > and then compare the samples using correlation function. > > > Before I start working on methodics and the programming itself (I have > > to admit that I only have remains of school knowledge in this area), I > > would like to ask if there are not any standard procedures, algorithms > > or software to start with.. > > Final aim should be to automate this testing using some script and not > > to click it by hand, but I'm interested also in directions to use some > > graphical audio processing software > > I prefer solution on linux platform, but it isn't prerequisite, Scilab/ > > Matlab is also OK. > > > Any ideas on where to start? > > > with best regards > > Actually there are quite a few open libraries for programming VoIP > applications. Have a look at EMIPLIB:http://research.edm.uhasselt.be/emiplib/emiplib.html > It makes use of JRTPLIB which is a library for the Real-time Transport > Protocol and RTCP. > RTCP will send out sender & receiver reports, and these will give you the > transport information you need.