"... In PAGE 15: ... Table 2 shows that av for Code 2 is larger than that for Code 3. In Table3 we provide an optimum BPSK ST code (Code 4), an optimum QPSK ST-2TCM code (Code 5) and an optimum 8PSK-2TCM code (Code 6) designed for a fast, at fading channel. Careful examination reveals that Code 6 is actually identical to Code 5 when viewed in terms of actual constellation symbols transmitted.... In PAGE 16: ... obtain extra coding gain by using higher order modulation in this case. Table3 also includes the corresponding theoretical coding gain = Qk2 (c;e) jck ? ekj2 calculated for the shortest error event path along which a diversity gain of l0 = min(j (c; e)j) = 2 is achieved. For Code 4, the shortest error event path has length l0 = 2 while we have l0 = 1 for Code 5 and Code 6.... ..."

"... In PAGE 16: ... Table 2 shows that av for Code 2 is larger than that for Code 3. In Table3 we show the best4 BPSK ST-TCM code (Code 4) we found, an optimum 4Note that Code 4 does not optimize the criterion in [1], but Code 4 is better for the nonasymptotic SNRs we consider.... In PAGE 17: ... Thus, we cannot obtain extra coding gain by using higher order modulation in this case. Table3 also includes the corresponding theoretical coding gain = Qk2 (c;e) jck ? ekj2 calculated for the shortest error event path along which a diversity gain of l0 = min(j (c; e)j) = 2 is achieved. For Code 4, the shortest error event path has length l0 = 2 while we have l0 = 1 for Code 5... ..."

"... In PAGE 15: ... Modulation p q code rate (time) BPSK 1 1 1 QPSK 2 2 1=2 8PSK 2 2 1=3 Table 1: Parameters for codes to be compared. In Table2 we provide three code examples for quasi-static at fading channels. Code 1 is the optimum (maximum diversity gain and maximum coding gain) BPSK ST-TCM code.... In PAGE 15: ... Optimality was veri ed using a complete search. The last two columns are just a permutation of the rst two columns in G in Table2 . The labeling across time follows a pattern similar to a repetition code.... In PAGE 15: ... It achieves the maximum coding gain (and diversity gain). Table2 also lists the corresponding theoretical coding gain . In Figure 3, we compare the performance of Code 1, Code 2 and Code 3 in a quasi-static at fading channel for cases with M = 2 and M = 4.... In PAGE 16: ... We obtain some information on the coding gain distribution by computing the coding gain averaged over all length-k error events 1 k 3. This quantity av is included in Table2 . Some justi cation and further discussion of av is given in [16].... In PAGE 16: ...ength-k error events 1 k 3. This quantity av is included in Table 2. Some justi cation and further discussion of av is given in [16]. Table2 shows that av for Code 2 is larger than that for Code 3. In Table 3 we show the best4 BPSK ST-TCM code (Code 4) we found, an optimum 4Note that Code 4 does not optimize the criterion in [1], but Code 4 is better for the nonasymptotic SNRs we consider.... ..."

"... In PAGE 19: ... A closed form pairwise SEP can be further derived [8]. The result of the pairwise SEP over two correlated fading channels is listed in Table1 . Note that the pairwise SEP for independent fading is the same as what is derived in [12].... In PAGE 19: ... Note that the pairwise SEP for independent fading is the same as what is derived in [12]. A closed form BEP for BPSK signal can be obtained by substituting in di = 1 and dj = ?1 in Table1 , and noting Es=Eb. The result is listed in Table 2 where Gb 4 = Eb=N0.... ..."

"... In PAGE 3: ... This may be unrealistic, but the effects of loss and reordering on the backbone do not have an effect on the performance of the link layer, which is the subject of this paper. Second, the error-prone channel is a bit-level Gilbert-Elliot model generated from a Rayleigh fading channel according to (Wang amp; Moayeri, 1995), assuming a raw channel capacity of 2 Mbps (parameters in Table1 ). Table 1 Parameters of the Gilbert-Elliot channel and of a corresponding Rayleigh fading channel (eG, eB: error probability in the good/bad channel state; mG, mB: average duration of the good/bad channel state).... ..."

"... In PAGE 5: ... Such results for a number of di#0Berent modulation methods and code rates for a Rayleigh fading channel are presented in #5B5#5D. Table1 shows these theoretical limits for BPSK modulation and soft decision decod- ing. We can see from Table 1 that for rate R =1=2 coding with BPSK modulation and soft decision sequential decoding is possible when E b =N 0 is above 8 dB.... In PAGE 5: ... Table 1 shows these theoretical limits for BPSK modulation and soft decision decod- ing. We can see from Table1 that for rate R =1=2 coding with BPSK modulation and soft decision sequential decoding is possible when E b =N 0 is above 8 dB. This limit has been veri#0Ced by simulations in #5B5#5D.... ..."

"... In PAGE 4: ....025, 0.054, 0.079 and 0.133, and for both the BER=1%- optimised speech and the BER=0.01%-optimised computer data transmission switching SNRs of Table1 in the absence of interference. It should be noted that during this evalu- ation of latency there was no limit imposed upon the bu er length.... In PAGE 7: ... The other advantage of de-correlating the fading is that the PDF of the instantaneous BER will have a lower variance and hence will be more concentrated around the average BER experienced by the source coded frames, since the amount of time spent in a deep fade is statistically speaking limited to a single burst, rather than in icting long strings of high-BER frames. This vindicates the decision to use the mean BER switching SNRs, from Table1 for these experi- ments rather than the peak BER switching SNRs. The proposed frequency hopping scheme was simulated as- suming M independent fading channels, with a frequency hop every TDD frame.... In PAGE 11: ...3. Performance Evaluation The switching SNRs to be used for the adaptive modu- lation scheme were the speech system apos;s optimised mean BER levels and the speech system apos;s peak BER levels from Table1 . We opted for employing the block code BCH(63,39,4), since it was the most robust code from the n = 63-bit long BCH code family, that would achieve the 41.... In PAGE 11: ... All users transmitting in the TDD frame had the same throughput requirements, av- erage channel SNR and fading statistics. Table 9 shows the DER, FER and TER for both the mean-BER and peak- BER switching SNRs of Table1 , in both the Slow and Fast Rayleigh fading channels, for various numbers of users at average channel SNRs, when all the users are generally achieving the desired error-rate and delay performance. Several observations may be made from Table 9: The DER with peak BER switching SNRs is higher than with mean BER switching SNRs.... In PAGE 15: ... Speci cally, if the SNR is slightly above a switching SNR, then an intolerable burst of errors may be experienced. Therefore, the peak BER switching SNRs that were portrayed in Table1 will be used in the following experiments. 11.... ..."

"... In PAGE 7: ... The switching levels used in these experiments are listed in Table 1. The mean BER and BPS performances were then nu- merically calculated utilizing Equations 11 and 12 and the switching levels listed in Table1 for speech and data trans- mission. The results are shown in Figure 7 for the COST 207 TU Rayleigh fading channel of Figure 2.... ..."

"... In PAGE 3: ... Average spectral distortion is used as performance measure. Table2 shows results for the average spectral distortion (SD). In this table, row marked as CELP shows performance results when a scalar quantization is applied to the LSP parameters.... In PAGE 3: ... Rows marked as COMQ-21, COMQ-12, COMQ-6 and COMQ-0 show performance results for our technique in which COMQ quantization codebooks aretrained at a CSNR of 21, 12, 6 and 0 dB, respectively. From Table2 it can be observed that in general perfor- mance results of experiment CELP are worse than perfor- mance of others experiment, and the performance difference grows with an increment of the noise in the channel. The exception is when the training of quantization codebooks is done under a very noisy channel condition.... In PAGE 3: ... For example, at a CSNR of 6 dB, experi- ment COMQ-6 gives the best performance results compared to the others experiments, for both channel models. From Table2 it is clear that COMQ-X coders outperform other considered coders, specially for a noisy channel. For exam- ple, for a Gaussian Channel at a CSNR of 6 dB, COMQ-12 gives a 0.... ..."