曝光台 注意防骗
网曝天猫店富美金盛家居专营店坑蒙拐骗欺诈消费者
executionsjbr each spch[]).
4.8.1 ERRORSI GNAL DERIVATION
For each delay, qgazn, qtaps[], synthspchfl, and s2[], the
vector error!] shall be generated by the operation defined by:
local-synthspch: =syntlzspch; {create a local copy
of the short term predictorjiher memory)
local-s2:=s2; {create a locar! copy uf the
long term predictor filter mernory}
FOR n:=h TO b+31 DO
BEGIN
surn:=O;
For each error[] and iresp[], the vector xcorr[] shall be
generated by the operation defined by:
FOR n:=O TO 31 DO
BEGIN
dot:=O;
FOR j:=n TO 31 DO dot:=dot+errur[b+j]*
iresp[j-nl;
xcorr[n]: =dot
END;
Note I.- The scalar b is the sample number corresponding
to the start of the currenr excitation frame within the
current speech frame. For rhe firsr, second, third, fourth and
jfth excitation frames of each speech frame, the value of b
equals 0, 32, 64, 96, and 128 respectively.
Note 2.- This operation is pe$ormed for each of the jive
excitation frames contained within a transmission frame.
4.8.4 PULSES ELECTION
For each xcorr[], ipwr[], and iresp[], the vectors posns[] and
qamp[] shaIl be generated by the operation defined by:
Annex ZO - Aeronautical Telecommunications Volume III
FOR pulse:=l TO 3 DO
BEGIN
rnax: = 0;
FOR n:=O TO 31 DO
IF (sqr(xcorr[n])>=mnx*ipwr[n]) and not
(n in posn-set) THEN BEGIN
ma: = sqr(xcorr[n])/;pwr[n];
pos:=n
END;
posn-set: =posn-wt+[pos];
posns[pulse]: =pos;
amp: = xcorr[pos]/ipwr[pos];
qarnp[pulse]: =quantamp(amp, pulse);
{Account for the effect of the quantized pulse
just calculated on the error to be minimized]
WHILE pulseenpulse DO
BEGIN
FOR n:=O TO pred(pos) DO
BEGIN
dot: =O;
FOR j:=O TO 31-pos DO dot:=dot+iresp[j]*
iresp[j+(pos-n)];
xcorr[n]: =xcorr/n-qamp *dot
END:
FOR n:= pus TO 31 DO
BEGIN
dot:=O;
FOR j:=O TO 31-n DO dot:=dot+iresp[j]*
iresp[j+(n-pos)];
xcorr[n]: =xcorr[n]-qamp *dot
END
END
END:
Note 1.- The second and third pulses selected in each
excitation frame are not allowed to occupy a position already
occupied by a previously selected pulse in that excitation
frame-
Note 2.- The function quantamp(amp,pulse) is defined in
4.8.4.1.
Note 3.- This operation is performed for each of thejve
excitation frames contained within a transmission frame.
Where invoked elsewhere, the function quantamp(amp,pulse)
shall be used to generate values in the vector qamp[] by the
operation defined by:
{3 bit quantizer initialization)
{4 bit quantizer initialization}
adapt[4,0]: =O. 75; adapt[4, I]: =0.8 75;
adapt[4,21:=0.875; adapt[4,3]:=0.875;
adap t[4,4]: = I. 0;
adapt[4,5]: = 1.25; &pt[4,6]: = 1.5; adapt[4,7]: =2.0;
quant[4,0]: =0.2582; quant[4,1]:=0.5224;
quant[4,2]: =O. 7996;
quant[4,3]: = 1-099; quant[4,4]:= 1.437;
quant[4,5]:=1.844; quant[4,6]:=2.401;
inv_quant[4,0]: =O. 1284; inv-quant[4,1]: =0.3881;
inv_quant[4,2]: =0.6568;
invLquant[4,3]: =0.9424; inv-quant[4,4,T: = 1.256;
inv_quant[4,5]: = 1.61 8; inv-quant[4,6]: =2.069;
inv_quant[4,7]: =2.733;
FUNCTION quantamp(lev:real, pulse:integer).'real;
VAR i,qbits : integer
sgn : -I.+];
BEGIN
lev: = lev*adjust[pulw]
IF pulsee3 THEN BEGIN qbits: -+ qlevels: = 7 END
ELSE BEGIN qbits:=3; qlevels: =3 END;
IF lev>O THEN sgn:=l ELSE sgn:=-I;
i:=O;
WHILX (abs( lev) >quant[qbits, i] *&act-leak) and
(i<qlevels) DO i:=succ(i);
quantamp:=inv-quant[qhits, i] *&act-leak*
sgn/adjust[pulse];
duct-leak: =EXP(O. 98 *WV(gfact-leak));
IF i=qlevels THEN
BEGIN
IF maxadapt THEN gfact_leak:=&act-leak"
adapt[qbits,qlevels]; rnaxadapt:=true
END ELSE maxadapt:=false;
&act-leak: =gfact-leak*adapt[qbits, i];
IF gfact_leak>512.0 THEN &act-leak: ~512.0E LSE
IF &act-lmkc0.5 THEN gfnct71eak:=0.5
END { of quantarnp. };
Note 1.- This operation is performedfor each of the jve
excitation franits contained within a transmission frame.
{Pulse amplitude quantizer initialization}
gfact-leak:= 1; mxdddpt: =false; Note 2.- The final, post-leakage pulse ainplitudes deveadjust[
l]:=l; adjusr[2]:=1.0/0.625; adjust[3]:=1.0/0.375; loped within the function quantampo are each quantized to a
Annex 10 - Aeronautical Telecommunications
ji-~ed~ ~albuaes ed on internal qunnrizatiorl thresholds; and then
couverted to correq?oizding 1in.e codes. The quantization and
li71e rode generation aspects of the function are, in efect, a
tuble lookup operation which can be visualized by referring to
中国航空网 www.aero.cn
航空翻译 www.aviation.cn
本文链接地址：
附件10--航空电信an10_v3_1ed(109)
