HTML5 Webook
96/182

same evaluation value if the same input is given.is study evaluated the speech quality using PESQ (ITU-T P.862), a media-layer model which is one of the objective speech quality evaluation methods.e evaluation conditions are as follows:*Speakers-Two men and two women-Fourteen speeches for each-Total of six minutes (two seconds between speeches)-16 kHz sampling*Connection-Directly connect to the combo jack of a smartphone through a cable from an audio interface.Figure 9 shows the devices used for sampling.Moreover, the “speech database for research of the Acoustical Society of Japan” was used as speaker data.3.3.3Result of speech quality evaluationWe evaluated the speech quality using PESQ (ITU-T P.862) for whether understanding is possible when con-necting to telephone network using a WINDS line and the voice communication system. Figure 10 shows the results of the speech quality evaluation. e vertical axis is PESQ values, and the horizontal axis is types of telephone ap-plications. Here, we respectively evaluated the cases of using a “disaster phone,” using “Zoiper” which is a SIP client, calling a xed telephone using “SkypeOut” from smartphone A and calling smartphone C using “LINE” from smartphone A, and we plotted the average values.e PESQ values of the “disaster phone” were 2.6 with the Speex8 k codec and 3.3 with the μ-law8 k codec, and the PESQ values of “Zoiper,” which is regarded as having good sound quality, were 2.7 with the Speex8 k codec and 3.1 with the μ-law8 k codec. erefore, large dierences are not shown from the results above.In addition, the PESQ value of “SkypeOut” is 3.4, and the PESQ value of “LINE” is 4.0, the largest value. ese are built by assuming primarily use under a high latency environment and the uctuation of packets on the internet paths.Although the evaluation methods are dierent, gener-ally, it is said that PESQ = 3.5 means that “90% of people” determine that “the said quality is better than usual (from the position of using phones routinely)”; PESQ = 3.1 means “80% of people” determine that “the said quality is better than usual (from the position of using phones routinely)”; and PESQ = 2.3 means that “40% of people” determine that “the said quality is better than usual (from the position of FiF8 Network configuration of voice communication systemFiF9 Devices for sampling3 Ultra-High-Speed Satellite Communication Technology92   Journal of the National Institute of Information and Communications Technology Vol. 64 No. 2 (2017)

元のページ  ../index.html#96

このブックを見る