V_ACTIVLEV Measure active speech level as in ITU-T P.56 [LEV,AF,FSO]=(sp,FS,MODE)
Usage: (1) lev=v_activlev(s,fs); % speech level in units of power
(2) db=v_activlev(s,fs,'d'); % speech level in dB
(3) s=v_activlev(s,fs,'n'); % normalize active level to 0 dB
(4) inc=10000; % you can process a long signal in chunks of length inc
fso=fs; % initialize fso to the sample frequency
for i=1:inc:ns % loop for each chunk
[lev,af,fso]=v_activlev(s(i:min(i+inc-1,length(s))),fso,'dz'); % include the 'z' option
end
lev=v_activlev([],fso); % finally omit the 'z' option to obtain the active level
% all other options, e.g. 'd', are preserved from the very first call
Inputs:
SP is the speech signal (with better than 20dB SNR)
FS is the sample frequency in Hz (see also FSO below)
MODE is a string containing a combination of the following:
0 - omit high pass filter completely (i.e. include DC)
3 - high pass filter at 30 Hz instead of 200 Hz (but allows mains hum to pass)
4 - high pass filter at 40 Hz instead of 200 Hz (but allows mains hum to pass)
1 - use cheybyshev 1 filter
2 - use chebyshev 2 filter (default)
e - use elliptic filter
h - omit low pass filter at 5.5, 12 or 18 kHz
w - use wideband filter frequencies: 70 Hz to 12 kHz
W - use ultra wideband filter frequencies: 30 Hz to 18 kHz
d - give outputs in dB rather than power
n - output a normalized speech signal as the first argument
N - output a normalized filtered speech signal as the first argument
l - give both active and long-term power levels
a - include A-weighting filter
i - include ITU-R-BS.468/ITU-T-J.16 weighting filter
z - do NOT zero-pad the signal by 0.35 s
Outputs:
If the "n" option is specified, a speech signal normalized to 0dB will be given as
the first output followed by the other outputs.
LEV gives the speech level in units of power (or dB if mode='d')
if mode='l' is specified, LEV is a row vector with the "long term
level" as its second element (this is just the mean power)
AF is the activity factor (or duty cycle) in the range 0 to 1
FSO is a structure of intermediate information that allows
you to process a speech signal in chunks (see usage example 4 above).
Processing a signal in chunks is slower and may not give identical results
because it will use slightly different thresholds. Note that you must use
the 'z' option for all calls except the last. When the FS input is a structure, all
options other than 'z' are taken from the FS structure rather than from the MODE input.
VAD is a boolean vector the same length as sp that acts as an approximate voice activity detector0001 function [lev,af,fso,vad]=v_activlev(sp,fs,mode) 0002 %V_ACTIVLEV Measure active speech level as in ITU-T P.56 [LEV,AF,FSO]=(sp,FS,MODE) 0003 % 0004 %Usage: (1) lev=v_activlev(s,fs); % speech level in units of power 0005 % (2) db=v_activlev(s,fs,'d'); % speech level in dB 0006 % (3) s=v_activlev(s,fs,'n'); % normalize active level to 0 dB 0007 % (4) inc=10000; % you can process a long signal in chunks of length inc 0008 % fso=fs; % initialize fso to the sample frequency 0009 % for i=1:inc:ns % loop for each chunk 0010 % [lev,af,fso]=v_activlev(s(i:min(i+inc-1,length(s))),fso,'dz'); % include the 'z' option 0011 % end 0012 % lev=v_activlev([],fso); % finally omit the 'z' option to obtain the active level 0013 % % all other options, e.g. 'd', are preserved from the very first call 0014 % 0015 % Inputs: 0016 % SP is the speech signal (with better than 20dB SNR) 0017 % FS is the sample frequency in Hz (see also FSO below) 0018 % MODE is a string containing a combination of the following: 0019 % 0 - omit high pass filter completely (i.e. include DC) 0020 % 3 - high pass filter at 30 Hz instead of 200 Hz (but allows mains hum to pass) 0021 % 4 - high pass filter at 40 Hz instead of 200 Hz (but allows mains hum to pass) 0022 % 1 - use cheybyshev 1 filter 0023 % 2 - use chebyshev 2 filter (default) 0024 % e - use elliptic filter 0025 % h - omit low pass filter at 5.5, 12 or 18 kHz 0026 % w - use wideband filter frequencies: 70 Hz to 12 kHz 0027 % W - use ultra wideband filter frequencies: 30 Hz to 18 kHz 0028 % d - give outputs in dB rather than power 0029 % n - output a normalized speech signal as the first argument 0030 % N - output a normalized filtered speech signal as the first argument 0031 % l - give both active and long-term power levels 0032 % a - include A-weighting filter 0033 % i - include ITU-R-BS.468/ITU-T-J.16 weighting filter 0034 % z - do NOT zero-pad the signal by 0.35 s 0035 % 0036 % Outputs: 0037 % If the "n" option is specified, a speech signal normalized to 0dB will be given as 0038 % the first output followed by the other outputs. 0039 % LEV gives the speech level in units of power (or dB if mode='d') 0040 % if mode='l' is specified, LEV is a row vector with the "long term 0041 % level" as its second element (this is just the mean power) 0042 % AF is the activity factor (or duty cycle) in the range 0 to 1 0043 % FSO is a structure of intermediate information that allows 0044 % you to process a speech signal in chunks (see usage example 4 above). 0045 % Processing a signal in chunks is slower and may not give identical results 0046 % because it will use slightly different thresholds. Note that you must use 0047 % the 'z' option for all calls except the last. When the FS input is a structure, all 0048 % options other than 'z' are taken from the FS structure rather than from the MODE input. 0049 % VAD is a boolean vector the same length as sp that acts as an approximate voice activity detector 0050 0051 % For completeness we list here the contents of the FSO structure: 0052 % 0053 % ffs : sample frequency 0054 % fmd : mode string 0055 % nh : hangover time in samples 0056 % ae : smoothing filter coefs 0057 % abl: HP filter numerator and denominator coefficient 0058 % bh : LP filter numerator coefficient 0059 % ah : LP filter denominator coefficients 0060 % ze : smoothing filter state 0061 % zl : HP filter state 0062 % zh : LP filter state 0063 % zx : hangover max filter state 0064 % emax : maximum envelope exponent + 1 0065 % ssq : signal sum of squares 0066 % ns : number of signal samples 0067 % ss : sum of speech samples (not actually used here) 0068 % kc : cumulative occupancy counts 0069 % aw : weighting filter denominator 0070 % bw : weighting filter numerator 0071 % zw : weighting filter state 0072 % 0073 % This routine implements "Method B" from [1],[2] to calculate the active 0074 % speech level which is defined to be the speech energy divided by the 0075 % duration of speech activity. Speech is designated as "active" based on an 0076 % adaptive threshold applied to the smoothed rectified speech signal. A 0077 % bandpass filter is first applied to the input speech whose -0.25 dB points 0078 % are at 200 Hz & 5.5 kHz by default but this can be changed to 70 Hz & 5.5 kHz 0079 % or to 30 Hz & 18 kHz by specifying the 'w' or 'W' options; these 0080 % correspond respectively to Annexes B and C in [2]. 0081 % 0082 % References: 0083 % [1] ITU-T. Objective measurement of active speech level. Recommendation P.56, Mar. 1993. 0084 % [2] ITU-T. Objective measurement of active speech level. Recommendation P.56, Dec. 2011. 0085 % 0086 % Revision History 0087 % 0088 % 2011-10-16 713 Initial version 0089 % 2012-11-02 2471 Correctd behaviour when there are zero or multiple solutions to A(l)=C(l)+M 0090 % 2012-11-19 2516 Modified comments 0091 % 2014-03-17 4346 Modified comments 0092 % 2014-07-09 4795 Improved plotting when no output arguments are specified 0093 % 2016-01-06 7336 Include wideband options from 2011 standard 0094 % 2017-02-07 9407 Added 'z' option and corrected output for an all-zero input signal 0095 % 2018-03-22 10436 Modified comments 0096 % 2018-09-21 10863 Renamed to start with "v_" 0097 % 2018-11-07 10988 Changed EOL style to native so checkout works on all machines 0098 % 2019-11-19 11190 Fixed error in calculating the activity factor; it now excludes the 0099 % zero-padding samples from the calculation. [thanks to Joe Begin] 0100 % 2022-01-18 Using the 'n' option now gives a signal with exactly 0dB active level (thanks 0101 % to Alastair Moore and Becky Vos). Fixed anomalous behavior when FS is a 0102 % structure and MODE input is omitted. 0103 % 2022-02-21 Fixed error in plot of power histogram that meant that the threshold and active 0104 % level lines were in the wrong position 0105 % 2022-02-28 Tidied up the labelling on the power histogram plot 0106 0107 % Copyright (C) Mike Brookes 2008-2019 0108 % Version: $Id: v_activlev.m 11190 2019-11-19 12:53:58Z dmb $ 0109 % 0110 % VOICEBOX is a MATLAB toolbox for speech processing. 0111 % Home page: http://www.ee.ic.ac.uk/hp/staff/dmb/voicebox/voicebox.html 0112 % 0113 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 0114 % This program is free software; you can redistribute it and/or modify 0115 % it under the terms of the GNU General Public License as published by 0116 % the Free Software Foundation; either version 2 of the License, or 0117 % (at your option) any later version. 0118 % 0119 % This program is distributed in the hope that it will be useful, 0120 % but WITHOUT ANY WARRANTY; without even the implied warranty of 0121 % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 0122 % GNU General Public License for more details. 0123 % 0124 % You can obtain a copy of the GNU General Public License from 0125 % http://www.gnu.org/copyleft/gpl.html or by writing to 0126 % Free Software Foundation, Inc.,675 Mass Ave, Cambridge, MA 02139, USA. 0127 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 0128 0129 persistent nbin thresh c25zp c15zp e5zp 0130 if isempty(nbin) 0131 nbin=20; % 60 dB range at 3dB per bin 0132 thresh=15.9; % threshold in dB 0133 % High pass s-domain zeros and poles of filters with passband ripple<0.25dB, stopband<-50dB, w0=1 0134 % w0=fzero(@ch2,0.5); [c2z,c2p,k]=cheby2(5,50,w0,'high','s'); 0135 % function v=ch2(w); [c2z,c2p,k]=cheby2(5,50,w,'high','s'); v= 20*log10(prod(abs(1i-c2z))/prod(abs(1i-c2p)))+0.25; 0136 c25zp=[0.37843443673309i 0.23388534441447i; -0.20640255179496+0.73942185906851i -0.54036889596392+0.45698784092898i]; 0137 c25zp=[[0; -0.66793268833792] c25zp conj(c25zp)]; 0138 % [c1z,c1p,c1k] = cheby1(5,0.25,1,'high','s'); 0139 c15zp=[-0.659002835294875+1.195798636925079i -0.123261821596263+0.947463030958881i]; 0140 c15zp=[zeros(1,5); -2.288586431066945 c15zp conj(c15zp)]; 0141 % [ez,ep,ek] = ellip(5,0.25,50,1,'high','s') 0142 e5zp=[0.406667680649209i 0.613849362744881i; -0.538736390607201+1.130245082677107i -0.092723126159100+0.958193646330194i]; 0143 e5zp=[[0; -1.964538608244084] e5zp conj(e5zp)]; 0144 % w=linspace(0.2,2,100); 0145 % figure(1); plot(w,20*log10(abs(freqs(real(poly(c15zp(1,:))),real(poly(c15zp(2,:))),w)))); title('Chebyshev 1'); 0146 % figure(2); plot(w,20*log10(abs(freqs(real(poly(c25zp(1,:))),real(poly(c25zp(2,:))),w)))); title('Chebyshev 2'); 0147 % figure(3); plot(w,20*log10(abs(freqs(real(poly(e5zp(1,:))),real(poly(e5zp(2,:))),w)))); title('Elliptic'); 0148 end 0149 if nargin<3 0150 mode=' '; 0151 end 0152 if ~isstruct(fs) % no state vector given 0153 fso.ffs=fs; % sample frequency 0154 ti=1/fs; 0155 g=exp(-ti/0.03); % pole position for envelope filter 0156 fso.ae=[1 -2*g g^2]/(1-g)^2; % envelope filter coefficients (DC gain = 1) 0157 fso.ze=zeros(2,1); 0158 fso.nh=ceil(0.2/ti)+1; % hangover time in samples 0159 fso.zx=-Inf; % initial value for v_maxfilt() 0160 fso.emax=-Inf; % maximum exponent 0161 fso.ns=0; 0162 fso.ssq=0; 0163 fso.ss=0; 0164 fso.sf=1; % scale factor when forming energy histogram 0165 fso.sfdb=0; % scale factor in dB 0166 fso.kc=zeros(nbin,1); % cumulative occupancy counts 0167 % s-plane zeros and poles of high pass 5'th order filter -0.25dB at w=1 and -50dB stopband 0168 if any(mode=='1') 0169 szp=c15zp; % Chebyshev 1 0170 elseif any(mode=='e') 0171 szp=e5zp; % Elliptic 0172 else 0173 szp=c25zp; % Chebyshev 2 0174 end 0175 flh=[200 5500]; % default frequency range +- 0.25 dB 0176 if any(mode=='w') 0177 flh=[70 12000]; % super-wideband (Annex B of [2]) 0178 elseif any(mode=='W') 0179 flh=[30 18000]; % full band (Annex C of [2]) 0180 end 0181 if any(mode=='3') 0182 flh(1)=30; % force a 30 Hz HPF cutoff 0183 end 0184 if any(mode=='4') 0185 flh(1)=40; % force a 40 Hz HPF cutoff 0186 end 0187 if any(mode=='r') % included for backward compatibility 0188 mode=['0h' mode]; % abolish both filters 0189 elseif fs<flh(2)*2.2 0190 mode=['h' mode]; % abolish lowpass filter at low sample rates 0191 end 0192 fso.fmd=mode; % save mode flags 0193 if all(mode~='0') % implement the HPF as biquads to avoid rounding errors 0194 zl=2./(1-szp*tan(flh(1)*pi/fs))-1; % Transform s-domain poles/zeros with bilinear transform 0195 abl=[ones(2,1) -zl(:,1) -2*real(zl(:,2:3)) abs(zl(:,2:3)).^2]; % biquad coefficients 0196 hfg=(abl*[1 -1 0 0 0 0]').*(abl*[1 0 -1 0 1 0]').*(abl*[1 0 0 -1 0 1]'); 0197 abl=abl(:,[1 2 1 3 5 1 4 6]); % reorder into biquads 0198 abl(1,1:2)= abl(1,1:2)*hfg(2)/hfg(1); % force Nyquist gain to equal 1 0199 fso.abl=abl; 0200 fso.zl=zeros(5,1); % space for HPF filter state 0201 end 0202 if all(mode~='h') 0203 zh=2./(szp/tan(flh(2)*pi/fs)-1)+1; % Transform s-domain poles/zeros with bilinear transform 0204 ah=real(poly(zh(2,:))); 0205 bh=real(poly(zh(1,:))); 0206 fso.bh=bh*sum(ah)/sum(bh); 0207 fso.ah=ah; 0208 fso.zh=zeros(5,1); 0209 end 0210 if any(mode=='a') 0211 [fso.bw,fso.aw]=v_stdspectrum(2,'z',fs); 0212 fso.zw=zeros(max(length(fso.bw),length(fso.aw))-1,1); 0213 elseif any(mode=='i') 0214 [fso.bw,fso.aw]=v_stdspectrum(8,'z',fs); 0215 fso.zw=zeros(max(length(fso.bw),length(fso.aw))-1,1); 0216 end 0217 else 0218 fso=fs; % use existing structure 0219 end 0220 md=fso.fmd; % md is used to determine all options except 'z' which uses mode 0221 nsp=length(sp); % original length of speech 0222 if all(mode~='z') 0223 nz=ceil(0.35*fso.ffs); % number of zeros to append 0224 sp=[sp(:);zeros(nz,1)]; 0225 else 0226 nz=0; 0227 end 0228 ns=length(sp); 0229 if ns % process this speech chunk 0230 % apply the input filters to the speech 0231 if all(md~='0') % implement the HPF as biquads to avoid rounding errors 0232 [sq,fso.zl(1)]=filter(fso.abl(1,1:2),fso.abl(2,1:2),sp(:),fso.zl(1)); % highpass filter: real pole/zero 0233 [sq,fso.zl(2:3)]=filter(fso.abl(1,3:5),fso.abl(2,3:5),sq(:),fso.zl(2:3)); % highpass filter: biquad 1 0234 [sq,fso.zl(4:5)]=filter(fso.abl(1,6:8),fso.abl(2,6:8),sq(:),fso.zl(4:5)); % highpass filter: biquad 2 0235 else 0236 sq=sp(:); 0237 end 0238 if all(md~='h') 0239 [sq,fso.zh]=filter(fso.bh,fso.ah,sq(:),fso.zh); % lowpass filter 0240 end 0241 if any(md=='a') || any(md=='i') 0242 [sq,fso.zw]=filter(fso.bw,fso.aw,sq(:),fso.zw); % weighting filter 0243 end 0244 fso.ns=fso.ns+ns; % count the number of speech samples 0245 fso.ss=fso.ss+sum(sq); % sum of speech samples (not used internally but available in fso output) 0246 ssq=sum(sq.*sq); % sum of squared new speech samples 0247 if ssq>0 && fso.ssq==0 % if these are the first non-zero speech samples 0248 fso.sf=ns/ssq; % scale factor to normalize the mean power of this chunk to 1 0249 fso.sfdb=10*log10(fso.sf); % scale factor in dB 0250 end 0251 fso.ssq=fso.ssq+ssq; % sum of all squared speech samples 0252 [s,fso.ze]=filter(1,fso.ae,abs(sq(:)),fso.ze); % envelope filter 0253 % we use the scale factor fso.sf in the following line to ensure that 0254 % the histogram binning is unchanged when sp is multiplied by a constant 0255 [qf,qe]=log2(fso.sf*s.^2); % take efficient log2 function, 2^qe is upper limit of bin 0256 qe(qf==0)=-Inf; % fix zero values 0257 [qe,qk,fso.zx]=v_maxfilt(qe,1,fso.nh,1,fso.zx); % apply the 0.2 second hangover 0258 oemax=fso.emax; % previous max value of qe+1 0259 fso.emax=max(oemax,max(qe)+1); % update the max value of qe+1 0260 if fso.emax==-Inf % if all samples so far are zero 0261 fso.kc(1)=fso.kc(1)+ns; 0262 else % we have had some non-zero samples 0263 qe=min(fso.emax-qe,nbin); % force in the range 1:nbin. Bin k has 2^(emax-k-1) <= s^2 < 2^(emax-k) 0264 wqe=ones(length(qe),1); 0265 % below: could use kc=cumsum(accumarray(qe,wqe,nbin)) but unsure about backwards compatibility 0266 kc=cumsum(full(sparse(qe,wqe,wqe,nbin,1))); % cumulative occupancy counts 0267 esh=fso.emax-oemax; % amount to shift down previous bin counts 0268 if esh<nbin-1 % if any of the previous bins are worth keeping 0269 kc(esh+1:nbin-1)=kc(esh+1:nbin-1)+fso.kc(1:nbin-esh-1); 0270 kc(nbin)=kc(nbin)+sum(fso.kc(nbin-esh:nbin)); 0271 else 0272 kc(nbin)=kc(nbin)+sum(fso.kc); % otherwise just add all old counts into the last (lowest) bin 0273 end 0274 fso.kc=kc; 0275 end 0276 end 0277 if fso.ns>nz % now calculate the output values 0278 if fso.ssq>0 0279 aj=10*log10(fso.ssq*(fso.kc).^(-1)); % aj(jj) = active level if all power is in first jj bins 0280 % following line is equivalent to cj=20*log10(sqrt(2).^(fso.emax-(1:nbin)-1)); 0281 cj=10*log10(2)*(fso.emax-(1:nbin)-1)-fso.sfdb; % lower limit of bin j in dB correcting for scale factor 0282 mj=aj'-cj-thresh; 0283 jj=find(mj(1:end-1)<0 & mj(2:end)>=0,1); % find +ve transition through threshold 0284 if isempty(jj) % if we never cross the threshold 0285 if mj(end)<=0 % if we end up below if 0286 jj=length(mj)-1; % take the threshold to be the bottom of the last (lowest) bin 0287 jf=1; 0288 else % if we are always above it 0289 jj=1; % take the threshold to be the bottom of the first (highest) bin 0290 jf=0; 0291 end 0292 else 0293 jf=1/(1-mj(jj+1)/mj(jj)); % fractional part of j using linear interpolation 0294 end 0295 lev=aj(jj)+jf*(aj(jj+1)-aj(jj)); % active level in decibels 0296 lp=10.^(lev/10); % active level in power 0297 if any(md=='d') % 'd' option -> output in dB 0298 lev=[lev 10*log10(fso.ssq/fso.ns)]; 0299 else % ~'d' option -> output in power 0300 lev=[lp fso.ssq/fso.ns]; 0301 end 0302 af=fso.ssq/((fso.ns-nz)*lp); 0303 else % if all samples are equal to zero 0304 af=0; 0305 if any(md=='d') % 'd' option -> output in dB 0306 lev=[-Inf -Inf]; % active level is 0 dB 0307 else % ~'d' option -> output in power 0308 lev=[0 0]; % active level is 0 power 0309 end 0310 end 0311 if all(md~='l') 0312 lev=lev(1); % only output the first element of lev unless 'l' option 0313 end 0314 end 0315 if nargout>3 0316 vad=v_maxfilt(s(1:nsp),1,fso.nh,1); 0317 vad=vad>(sqrt(lp)/10^(thresh/20)); 0318 end 0319 if ~nargout 0320 vad=v_maxfilt(s,1,fso.nh,1); 0321 vad=vad>(sqrt(lp)/10^(thresh/20)); 0322 levdb=10*log10(lp); 0323 clf; 0324 subplot(2,2,[1 2]); 0325 tax=(1:ns)/fso.ffs; 0326 plot(tax,sp,'-y',tax,s,'-r',tax,(vad>0)*sqrt(lp),'-b'); 0327 xlabel('Time (s)'); 0328 title(sprintf('Active Level = %.1f dB, Activity = %.0f%% (ITU-T P.56)',levdb,100*af)); 0329 v_axisenlarge([-1 -1 -1.4 -1.05]); 0330 if nz>0 0331 hold on 0332 ylim=get(gca,'ylim'); 0333 plot(tax(end-nz)*[1 1],ylim,':k'); 0334 hold off 0335 legend('Signal','Smoothed envelope','VAD * Active-Level','Zero-padding','Location','SouthEast'); 0336 else 0337 legend('Signal','Smoothed envelope','VAD * Active-Level','Location','SouthEast'); 0338 end 0339 ylabel('Amplitude'); 0340 % lower right plot: diagram of threshold and active level calculation 0341 subplot(2,2,4); 0342 plot(cj,repmat(levdb,nbin,1),'k:',cj,aj(:),'-b',cj,cj,'-r',levdb-thresh*ones(1,2),[levdb-thresh levdb],'-r'); 0343 xlabel('Threshold (dB)'); 0344 ylabel('Active Level (dB)'); 0345 legend('Active Level (AL)','Speech>Thresh','Threshold','Location','NorthWest'); 0346 v_texthvc(levdb-thresh,levdb-0.5*thresh,sprintf('%.1f dB ',thresh),'rmr'); 0347 v_axisenlarge([-1 -1.05]); 0348 ylim=get(gca,'ylim'); 0349 set(gca,'ylim',[levdb-1.2*thresh max(ylim(2),levdb+1.9*thresh)]); 0350 kch=filter([1 -1],1,kc); 0351 % lower left plot: power histogram 0352 subplot(2,2,3); 0353 bar(5*log10(2)+cj(end:-1:1),kch(end:-1:1)*100/kc(end)); 0354 set(gca,'xlim',[cj(end) cj(1)+10*log10(2)]); 0355 ylim=get(gca,'ylim'); 0356 hold on 0357 plot(levdb([1 1]),ylim,'k:',levdb([1 1])-thresh,ylim,'r:'); 0358 hold off 0359 v_texthvc(levdb(1),ylim(2),'AL','rtk'); 0360 v_texthvc(levdb(1)-thresh,ylim(2),sprintf('Threshold'),'rtr'); 0361 xlabel('Frame power (dB)') 0362 ylabel('% frames'); 0363 elseif any(md=='n') || any(md=='N') % output normalized speech waveform 0364 fsx=fso; % shift along other outputs 0365 fso=af; 0366 af=lev; 0367 if any(md=='n') 0368 sq=sp; % 'n' -> use unfiltered speech 0369 end 0370 if fsx.ns>0 && fsx.ssq>0 % if there has been any non-zero speech 0371 lev=sq(1:nsp)/sqrt(lp); 0372 else 0373 lev=sq(1:nsp); 0374 end 0375 end