Roman; I recently found a useful document on MOS measurement published by ETSI/3GPP. It contains MOS score table with four kinds of network parameters, bit error rate(or radio condition, 10E-2, 5x10E-4), packet loss rate (0%, 3%), codec bandwidth (6.7, 12.2, 12.65, 15.85kbps) and delay (300ms, 500ms) disturbance. MOS value is provided for every combination of network paramete. Group of hunam beings are involved in to score the perceptual quality of VoIP call. Quality is categorized into five; Voice quality, Understanding, Interaction and perception and Global quality.
You may download the document for free from http://portal.etsi.org/Portal_Common/home.asp --> Service Index --> ETSI standard --> Publication Download and search for "3gpp tr 26.935". You may need to register yourself. Thanks and hope it helps Joonbum ---------------------------------------------------------------------- Message: 1 Date: Sat, 08 Dec 2007 10:37:18 +0500 From: Roman Imankulov <[EMAIL PROTECTED]> Subject: Re: [Linphone-users] MOS reporting To: [email protected] Message-ID: <[EMAIL PROTECTED]> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Hi, > > Does Linphone have a feature of calculating MOS (Measn Openion Score) of a > voice call? If yes, how the score is reported, RTCP, SNMP or ...? > As I know there is no satisfactory way to calculate MOS on the fly. * Originally proposed method (see ITU-T Rec. P.800) requires absolutely unsuitable conditions (such as specially-equipped studios, etc) * There is also PESQ (ITU-T Rec. P.862) which requires both reference and degraded speech samples (as I suppose, this is also unsuitable for the real-time quality estimation). * Last chance is an E-model (ITU-T Rec. G.107) which originally was proposed as quality estimation model for the circuit switching networks. Some of parameters of this model (such as one-way delay) cannot be calculated without additional measurements. I'd like to hear if somebody knows another ways to calculate speech quality or knows how to implement in real-time VoIP application some of methods proposed below. Also I'm not sure but it seems that PESQ and E-model are patented. -- WBR, Roman Imankulov [EMAIL PROTECTED]
_______________________________________________ Linphone-users mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/linphone-users
