vcdet4st.m

来自「这是一个用于语音信号处理的工具箱」· M 代码 · 共 183 行

disp(' ');
disp('SCRIPT:   vcdet4st.m ***********************************************');
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%	
%	Voiced Consonant Detection Program
% 	jmw 	
%	1/6/94
%	1/3/94
%	12/30/93
%	
% playing with Weinstein, et al., algorithm for parsing based
% upon RMS values of BP filters
% gain calculated from sqrt of res NRG (this feature removed)
%
%	THIS VERSION ALSO ADDS: multiple thresholds to better determine 
%	accurate v/c transition location...
%	
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

		PLT = 1;	% flag to plot results

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

file_string = sprintf('temp/%s.mat', name);
s=sprintf('loading ./%s from hard disk ...',file_string);
disp(s);
s=sprintf('load %s', file_string);
eval(s);
signal = eval(name);

file_string = sprintf('temp/%s_Data.mat', name);
s=sprintf('loading ./%s from hard disk ...',file_string);
disp(s);
s=sprintf('load %s', file_string);
eval(s);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% compute freq response

N=256;
start1 = 2;
stop1 = 52;
start2 = 53;
stop2=N;

[m,n]=size(cofa);
vc = zeros(m,1);
e1 = zeros(m,1);
e2 = zeros(m,1);
gain = zeros(m,1);

for i=1:m-1,

	if(VUS_voicetype(i)=='v' | VUS_voicetype(i+1)=='v'   )
		a = range(i,1);
		b = range(i,2);
	
		%res_nrg = residue(a:b) * residue(a:b)' ;
		%gain(i) =	sqrt(res_nrg); % removed jmw 9/30/94

		gain(i) = 1;	

		[h,w]=freqz(gain(i),cofa(i,:),N);
		mag=abs(h);
	
		e1(i)=sqrt(mag(start1:stop1)' * mag(start1:stop1));
		e2(i) = sqrt(mag(start2:stop2)' * mag(start2:stop2));
		vc(i) = e1(i) /e2(i);
	else
		;
	end;
end;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% smooth vc ratio

MFO = 5; % swiched back from 7 to 5 2/21/94	% median filter order
s=sprintf('smoothing vc ratio with median filter of order %d ...', MFO);
disp(s);
vc=median1(vc,MFO);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% calculate (voiced) consonant score

vc_thresh_LOW = 8;			% thresholds - empirically determined
vc_thresh_HI  = 18;
R = vc_thresh_HI - vc_thresh_LOW;		% R= "range"
vc_score = zeros(1,m);

for i=1:m,
	if (VUS_voicetype(i)=='b' | VUS_voicetype(i)=='v')
		if (vc(i) >=  vc_thresh_HI)
			vc_score(i) = 1;
		elseif (vc(i) > vc_thresh_LOW)
			vc_score(i) = (vc(i) - vc_thresh_LOW) / R;
		else
			vc_score(i) = 0;
		end;
		
	end;
end;
		
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

titles = 1; % turn off for diss figs

if (PLT)
	disp('plotting results ...');

	h=gcf;
	figure(h);
	clf;
	
	subplot(311);
	plot(signal);
	grid;
	axis_1=axis;
	axis(axis_1);
	s=sprintf('%s', name);
	if (titles)
		title(s);
	end;

	subplot(312);
	stairs(range(:,1),vc);
	grid;
	axis([axis_1(1) axis_1(2) 0 40]);
s=sprintf('smoothed vc ratio (0-1000)/(1000-5000) - order %d median filt', MFO);
	hold on;
	%plot(range(:,1), ( vc_thresh_HI * ones(1,m)), ':');
	%plot(range(:,1), ( vc_thresh_LOW * ones(1,m)), ':');
	plot(( vc_thresh_HI * ones(1,axis_1(2))), '--');
	plot( ( vc_thresh_LOW * ones(1,axis_1(2))), '--');
	hold off;
	if (titles)
		title(s);
	end;

	subplot(313);
	stairs(range(:,1),vc_score);
	grid on;
	axis([axis_1(1) axis_1(2) -0.5 1.5]);
	if (titles)
		title('(Voiced) Consonant Score (voiced segments only)');
	end;

	drawnow;
end;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% save to disk

s=sprintf('saving ./%s_VCscore to disk ...', name);
disp(s);
s=sprintf('save temp/%s_VCscore vc_score', name);
eval(s);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% clean up time

clear MFO                 e1                  res_nrg             
clear N                   e2                  residue             
clear PLT                 file_string         signal              
clear R                   gain                start1              
clear VUS_voicetype       h                   start2              
clear a                   i                   stop1               
clear axis_1              m                   stop2               
clear b                   mag                 vc                  
clear vc_score             n                   vc_score            
clear vc_thresh_HI                             voicetype           
clear vc_thresh_LOW        power               w                   
clear cofa                range titles

s=sprintf('clear %s',name);
eval(s);
clear s