"... In PAGE 8: ... Using this definition, we can say that our system is performing in real time when, for example, the system is able to process more than 32 KB per second when streaming 32 KHz 8-bit audio. Table2 shows the expected performance for a real-time system. Table 2 List of frequencies and sample sizes with the corresponding data rate and needed performance for real time limits ... ..."

"... In PAGE 7: ... Unfortunately, the present and potential experimental bounds on n28S n00 n06 P n00 n29 couplings are exceptionally poor or non-existent from measurements of the n19, n1a and a 1 modes. Table2 : Limits on n03 apos;s from n1c ! n1an17; a 1 n17 for Real Coupling Constants fn1a n00 ;n1a + g mode fa n00 1 ;a + 1 g mode At M Z 10, or 4 GeV At M Z 10, or 4 GeV V + A, for n18 A 0:006 0:0012 0:010 0:0018 S + P , for n03 310GeV 1; 700 64 350 S n00 P , for n28n03n29 2 n2811GeV n29 2 n2825n29 2 n284n29 2 n287n29 2 ; n2810n29 2 f M + f E , for n03 210GeV 1; 200 280 1; 500 f M n00 f E , for n28n03n29 2 n289GeV n29 2 n2820n29 2 n2810n29 2 n2824n29 2 For the n1a and a 1 modes, the T + + T + 5 coupling is equivalent to the V n00 A coupling; and T + n00 T + 5 is equivalent to V + A. Table 3: Limits on n03 apos;s from n1c ! n19n17: From n18 n19 : From n00n28n1c ! n19n17n29 jg i =g L j 2 jg i =g L j 2 2Ren28g L n03 g i n29 V + A, for n18 n19 0:015; 0:004; 0:009 0:014 S + P; T + + T 5 + , forn03 127GeV S n00 P; T + n00 T 5 + ,forn28n03n29 2 n2810GeV n29 2 ; n2821GeV n29 2 ; n2813GeV n29 2 n28n3c 1GeV n29 2 S n00 + P n00 , forn03 n3c 1GeV S n00 n00 P n00 ,forn28n03n29 2 n28n3c 1GeV n29 2 ; n281:6GeV n29 2 ; n281GeV n29 2 n28n3c 1GeV n29... ..."

"... In PAGE 8: ....51}. However in realistic situations if the performance differences are high, the worst classifiers are rejected and consequently the final differences are never that extreme. The results in Table2 showing extremely small absolute discrepancies from the real limits confirm our expectations. The limits component of the RE measure can be therefore considered as a fast and accurate estimation of the true MV error limits.... ..."

"... In PAGE 8: ....51}. However in realistic situations if the performance differences are high, the worst classifiers are rejected and consequently the final differences are never that extreme. The results in Table2 showing extremely small absolute discrepancies from the real limits confirm our expectations. The limits component of the RE measure can be therefore considered as a fast and accurate estimation of the true MV error limits.... ..."

"... In PAGE 7: ... The larger the number of supporting model forecasts (SMF) is, the more convincing is the interpretation. The second column of Table2 shows the percentages of the SD observations receiving support of at least one, two, three or four SMF. A fraction of the highly asymmetric SD observations can be concluded to indicate a real atmospheric asymmetry.... In PAGE 7: ... 5.2 Statistical significance of the support It is obvious that the procedure described above would result in non-zero percentages in Table2 even in a case of artificial SD observations being definitely independent of real atmo- spheric conditions. In other words, some unknown percent- age of the artificial observations would accidentally receive support from the NWP model forecasts and would further be concluded to show an asymmetric meteorological signal.... In PAGE 7: ... In other words, some unknown percent- age of the artificial observations would accidentally receive support from the NWP model forecasts and would further be concluded to show an asymmetric meteorological signal. Therefore, confidence limits to the percentages in Table2 are estimated in the following way: A sample of one hundred sets of gaussian random numbers is constructed. Each set consists of 296 604 values being definitely not related to real www.... In PAGE 8: ... The conclusions of the experiment will be sensitive only to the order at which the random numbers happen to occur. The resulting 95% and 99% confidence limits are included in the third and fourth columns of Table2 for each level of SMF. The percentage of highly asymmetric SD observations indicating real atmospheric asymmetry exceeds the 99% con- fidence limit at all SMF levels.... ..."

"... In PAGE 60: ...0024, a standard deviation ( ) of 4.2697 ( Table14 , reciprocal of the Btrans parameter), and an index value (-X ) of 6.1622 (Table14).... ..."