June 15, 2014
June 15, 2014
June 18, 2014
Electrical and Computer
24.104.1 - 24.104.11
A Speech Quality and Intelligibility Assessment Project Using Google VoiceAbstractIn this paper, a speech quality and intelligibility assessment project using Google Voice isintroduced. Speech quality and intelligibility assessment is an essential topic in any audio andspeech signal processing course, which is widely offered as a senior elective or graduate levelcourse in the electrical engineering curriculum. Speech quality and intelligibility assessment hasbecome an extremely important topic due to the increasing demand in areas such as digitalprotocol analysis, Dolby-dts compliance analysis, audio device testing, etc.Traditionally, one of the biggest challenges in studying speech quality and intelligibilityassessment is the availability of real-life data. The voicemail feature of the Google Voice serviceallows users to download their own voicemails which are real Voice-over-Internet-Protocol(VoIP) speech. The voicemails are sampled at 11025Hz and compressed as MPEG audio layerIII (mp3). Google Voice also offers automatic transcription of the voicemails, which offers greatmaterial for students to examine the effects of vocabulary and contextual information on theintelligibility of speech.The speech quality and intelligibility assessment project introduced here includes three stages. Instage 1, students create a speech data bank using their own voice. Each student creates two setsof recordings. For the first set of recordings, they read the famous “rainbow passage” usingcontinuous speech. For the second set of recordings, they read a randomly selected group ofwords in the classic Modified Rhyme Test (MRT) using discrete speech. For each reading, theysimultaneously record their speech using a close talking microphone on their own PC, and at thesame time through a cellphone and record their speech using Google Voice’s voicemail system.After uploading their own recordings onto a commonly shared webpage (e.g. Blackboard’s ‘fileexchange’ page), students have access to each other’s recordings, which comprise a completespeech database including both the VoIP degraded speech and the clean reference speech. Instage 2, students subjectively assess the speech quality and speech intelligibility of eachrecording using the Mean Opinion Score (MOS) system. In stage 3, students create their ownMatlab programs to assess the speech quality and speech intelligibility of the recordingsobjectively. The algorithm used is similar to the Perceptual Evaluation of Speech Quality(PESQ) standard which is a legacy industry standard for objective voice quality testing.Students’ feedback in terms of their experience using Google Voice and their learningexperience are presented. The instructor’s own observations and conclusions, and future plan arepresented at the end.
Yu, Y. (2014, June), A Speech Quality and Intelligibility Assessment Project Using Google Voice Paper presented at 2014 ASEE Annual Conference & Exposition, Indianapolis, Indiana. 10.18260/1-2--19996
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 2014 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015