Loss Rate (LR)

As shown in all the Figures, the most dominant effect on the quality is due to the LR. Speech quality degrades severely when LR increases for constant values of the other parameters. As shown in Figure 7.1(a), for PCM codec at PI=20ms, the quality degrades from 3.7 to 1.5 (about 2.2 points) and from 4.1 to 2.2 (about 1.9 points) when LR increases from 5 to 40 % for CLP=1 and 5 respectively. The same results is qualitatively the same for the other codecs, knowing that PCM gives the best quality for the same values of the other parameters. From Figure 7.2, we can see that the effect of CLP and PI is more important in the situation where LR=10 % than that where LR=5 %. In addition, we can get the same quality by changing these two parameters if the LR changes. For example, for CLP=2 using ADPCM codec, the quality when LR=5 % and PI=80ms is the same as LR=10% and PI=20ms (about 3 points). The same observation is for CLP. In addition, if we have poor network conditions, by changing the codec, the quality can be improved. A difference of 0.5 point is generally observed when changing from ADPCM to PCM. When there are lost packets in the speech flows there are two effects that degrade speech quality. The first is that lost parts of some words occur and parts of speech become incomprehensible. The other effect is due to the encoding and decoding algorithms. Some encoding algorithms use the information of the past speech frames to encode the current. If one frame is lost many frames cannot be decoded and the effect of loss becomes more annoying than for other codecs that do not use temporal compression. The effect of LR has been studied in some related works in the literature, namely in [28,61].