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
c) at 4 600 Hz and above, a gain of -75 dB or lower.
Note I.- In addition to the listed spectral characteristics,
additional jilzering appropriate to the particular equipment
installation should be provided in the low-end response
characteristics of the vocoder so as to mitigate the effects of
power supply noise pickup (e.g. 400 Hz for AES installations
and 50/60 Hz for GES installations).
Note 2.- The input to the encoding algorithm is dejned
in 3.2 to be quantized audio samples with at least 12 bits of
trsolution. Consequently, all required spectral characteristics
at the point of sampling are defined with this in mind. The use
of u-law or A-law companded audio in conjr~nctionw ith borh
reduced resolution (e.g. 8 bits) and appropriately relaxed
transition-band and stop-band gain characteristics is permitted
provided that the quantized samples are decompanded
(linearised) prior to their presentation to the algorithm.
* All figures are located at the end of this chapter.
9/11/95 198 ANNEX 10 - VOLUME 111
Part I Annex 10 - Aeronautical Telecommunications
3.2 Sampling, quantization
and scaling
Audio which exhibits the spectral characteristics defined in 3.1
shall be sampled at a continuous 8 kHz rate, and the resultant
series of samples shall be quantized to an equivalent timeordered
series of linear (i.e. non-companded) binary values.
Each quantization operation shall be performed with a precision
of at least 12 bits and a linearity of at least 10 bits. The
maximum dynamic range of this series shall be scaled up or
down in magnitude as appropriate so that it is tightly bounded
by the floating-point range of 4 095.0 to -4 096.0 without
exceeding it. The resultant time-ordered series of scaled values
shall be presented to the segmentation and windowing logic
defined in 4.1.
Note.-- The above bounding range of scaled values rS
indicative of a fied-point signed value of 13 bits zn length.
Scaling of the time-ordered series of binary values to the
indicated numeric range rs required in order to take muximum
advaiztage qf the ir~ternaln umeric raoge of tlze ulgorithm.
4. ENCODING PROCESS DEFINITION
4.1 Segmentation and windowing
The time-ordered stream of inpot samples shall be subdivided
into time-ordered segments termed segspchl], each comprising
256 time-ordered input samples (32 ms). Each segsprh[] shall
be so constructed that the first and last 96 samples (12 ms)
contained within it overlap with the immediately previous and
subsequent segspch[]. Each segspch[] shall be windowed with
a Hamming window vector hamwindow[] defined by:
FOR n:=O TO 255 DO hamwindow[n]:=
0.54-0.46*COS(2*3,1415927*d55);
and a windowing operation defined by:
FOR n:=O TO 255 DO windspch[n]:=
segspch[n]*hamwindow[n];.
An additional time-ordered vector spchl] corresponding to the
central 160 samples of each segspchl] vector shalI be
generated by the operation defined by:
FOR n:=O TO 159 DO spch[]:=segspch[n+48];.
4.2 Au tocorrelation
For each windspch[], an eleven element autocorrelation vector
corr[] shall be generated by the operation defined by:
FOR i:=O TO 10 DO
BEGIN
dot: =o;
FOR n:=O TO 255.; DO dor:=dor+windspch[rz]
*windspch[n+i];
corr[i ]:=dot
END;
4.3 Durbin's recursion
For each corr[], a partial correlation coefficient vector
yarcor[] shall be generated by the operation defined by:
IF corr[0]=0 THEN FOR i:-l TO 10 DO parcor[i]:=O;
ELSE
BEGIN
parcor[l/:=corr[l]/corr[O];
raps[l]:=parcor[l];
error:= ( I-sq r(parcor[l])) *corr[O];
FOR i:=2 TO 10 DO
BEGIN
FOR ;:=I TO pred(i) DO alpha[~]:=taps[j];
parc\or[i]:=corr[i];
FOR J:= I TO pred(i) DO parcor[i]:=parcor[i]-
alpha[j] *corr[i-j];
parcor[i]:=parcor[i]/error;
taps[i]: =par-cor[i];
FOR j:=I TO pred(i) DO taps[j]:=alpha[J]-
parcor[i]*aZplza[i-j];
error:=(]- sqr(parcor[i])")e rror
END
END;
4.4 Quantization of the partial
correlation coefficients
For each parcor[], the ten partial correlation coefficients
contained within shall be quantized and stored in vector
qparcor[] which is generated by the operation defined by:
{Quantizer iizitialization procedure)
{No. of quantizer levels for each coeficient)
Annex 10 - Aeronautical Telecommunications Volume 111
FOR i:=1 TO 10 DO
FOR j:=0 TO pred(plevel[i]) DO
parql[i,j]:=pardl[i,pred(jl]+pardl[i,j]n;
{Partial correlation coeficient quantizer]
FOR i:=l TO 10-DO
BEGIN
j:=O;
WHlLE (parcor[i]>pardl[i,j]a nd
(j<pred(plevel[i])) DO j:=succ(j);
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(107)
