mft6.m

这是一个用于语音信号处理的工具箱

disp(' ');
disp('SCRIPT:   mft6.m ***********************************************');
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% 	
%	McCandless Formant Tracker attempt #6 
%	This is the final version used in jmw's dissertation
% 
%	jmw 
%	2/15/94
%	12/21/93
%
% 	uses last formant freqs for new estimates. 
% 	AND starts at highest power and works forward andbackward
%
% 	this version saves formant tracks to hard disk (AND AMPLITUDES!)
%
%	initial values are for MALE voices ONLY.
%	See McCandless Paper for initial estimates for female voices ...
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

		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);

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

% do the entire sequence of steps for each v  frame

VERB1 = 0;		% verbosity flags
VERB2 = 0;

[m,n]=size(cofa);	% get size of data
f1=zeros(1,m);
f2 = f1;
f3 = f1;
f4 = f1;

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

% find valid frame with highest power
[sort_power,power_index]=sort(power);

target_frame = 0;

for j=1:10,
	if(target_frame==0 & VUS_voicetype(power_index(m-j+1))=='v')
		target_frame = power_index(m-j+1);
	end;
end;

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

disp('finding formants ...');

for i=target_frame:-1:1,	% begin main loop backwards

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

allslotsfilled = 0;
MAXPASSES = 30;
r = 1.0;
passcnt = 1;		% count number of passes

while( ~allslotsfilled & (passcnt<=MAXPASSES) & (r >= 0.88) ),

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

% STEP #1: find peaks for all frames that are 'v'

p   = [ 0 0 0 0 ];	% allocate mem for peaks for curremt frame
amp = [ 0 0 0 0 ];	% allocate mem for amplitudes of pks for curremt frame
	
if(VUS_voicetype(i)=='v') 	% find peaks if v
	
	% adjust radius depending upon pass number
	if (passcnt==1)
		r = 1.0;
	elseif (passcnt ==2)
		r = 0.98;
	else
		r=r-0.004;
	end;

	[p,a]=fpeak4(cofa(i,:),r);
	
	if (VERB1)		
		s=sprintf('pass %d ********  frame %d ***** (%d) **********',...
			passcnt,i,range(i,1));
		disp(s);
	end;
	
	if (VERB2 )

		disp('p and a');
		disp([p' a' ]);
	end;
end;

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

% STEP #2: fill slots with peaks using estimates as a guide

f_initial = [ 0 0 0 0 ];	% allocate mem

if(VUS_voicetype(i)=='v') 	% fill slots if v
	
	
	% check if previous frame has valid formant values
	f_prev = [f1(i+1) f2(i+1) f3(i+1) 3200];
	f_prev_zero_index = find(f_prev == 0);
	f_prev_valid = isempty(f_prev_zero_index);

	if (f_prev_valid)
		f_initial = fillslot(p, f_prev);
	else
		f_initial = fillslot(p);
	end;	

	if (VERB2)
		disp('f_initial');
		disp([f_initial']);
	end;
end;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% STEP #3: remove duplicates using estimates as a guide

if (VUS_voicetype(i)=='v') 	% check slots if v
	
	% check if previous frame has valid formant freqs
	if (f_prev_valid)
		f_nodup = rmdup4a(f_initial,f_prev);
	else
		f_nodup = rmdup4a(f_initial);
	end;	
	
	if (VERB2)
		disp('f_nodup');
		disp([f_nodup']);
	end;
end;

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

% STEP 4: Deal with unassigned peaks

if (VUS_voicetype(i)=='v')  % check unused pks if v
	
	f_ed1 = redounas(f_nodup, p, a);
	
	if (VERB2)
		disp('f_ed1');
		disp(f_ed1');
	end;

end;

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

% save results in big array

if(VUS_voicetype(i)=='v')  	
	f1(i) = f_ed1(1);
	f2(i) = f_ed1(2);
	f3(i) = f_ed1(3);
	f4(i) = f_ed1(4);	
end;

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

% check if all slots filled if v 

if(VUS_voicetype(i)=='v')  % check  if v
	zeroindex = [];
	fcheck = [f1(i) f2(i) f3(i) ]; % dont care about f4
	zeroindex = find( fcheck == 0);
	allslotsfilled = isempty(zeroindex);
	if (VERB1 & allslotsfilled)
		disp('ALL SLOTS FILLED');
	elseif (VERB1 & ~allslotsfilled)
		s=sprintf('formant %d NOT FILLED', zeroindex);
		disp(s);
	end;
	passcnt = passcnt+1;
else
	allslotsfilled = 1;		
end;


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

end;	% end while loop

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

end;	% end main loop backwards

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

for i=target_frame+1:m,	% begin main loop forwards

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

allslotsfilled = 0;
MAXPASSES = 30;
r = 1.0;
passcnt = 1;		% count number of passes

while( ~allslotsfilled & (passcnt<=MAXPASSES) & (r >= 0.88) ),

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

% STEP #1: find peaks for all frames that are 'v'

p   = [ 0 0 0 0 ];	% allocate mem for peaks for current frame
amp = [ 0 0 0 0 ];	% allocate mem for amplitudes of pks for curremt frame
	
if(VUS_voicetype(i)=='v') 	% find peaks if v
	
	% adjust radius depending upon pass number
	if (passcnt==1)
		r = 1.0;
	elseif (passcnt ==2)
		r = 0.98;
	else
		r=r-0.004;
	end;

	[p,a]=fpeak4(cofa(i,:),r);
			
	if (VERB1)
		s=sprintf('pass %d ********  frame %d ***** (%d) **********',...
			passcnt,i,range(i,1));
		disp(s);
	end;

	if (VERB2 )

		disp('p and a');
		disp([p' a' ]);
	end;
end;

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

% STEP #2: fill slots with peaks using estimates as a guide

f_initial = [ 0 0 0 0 ];	% allocate mem

if(VUS_voicetype(i)=='v') 	% fill slots if v
	
	
	% check if previous frame has valid formant values
	f_prev = [f1(i-1) f2(i-1) f3(i-1) 3200];
	f_prev_zero_index = find(f_prev == 0);
	f_prev_valid = isempty(f_prev_zero_index);

	if (f_prev_valid)
		f_initial = fillslot(p, f_prev);
	else
		f_initial = fillslot(p);
	end;	

	if (VERB2)
		disp('f_initial');
		disp([f_initial']);
	end;
end;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% STEP #3: remove duplicates using estimates as a guide

if (VUS_voicetype(i)=='v') 	% check slots if v
	
	% check if previous frame has valid formant freqs
	if (f_prev_valid)
		f_nodup = rmdup4a(f_initial,f_prev);
	else
		f_nodup = rmdup4a(f_initial);
	end;	
	
	if (VERB2)
		disp('f_nodup');
		disp([f_nodup']);
	end;
end;

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

% STEP 4: Deal with unassigned peaks

if (VUS_voicetype(i)=='v')  % check unused pks if v
	
	f_ed1 = redounas(f_nodup, p, a);
	
	if (VERB2)
		disp('f_ed1');
		disp(f_ed1');
	end;
end;

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

% save results in big array

if(VUS_voicetype(i)=='v')  	
	f1(i) = f_ed1(1);
	f2(i) = f_ed1(2);
	f3(i) = f_ed1(3);
	f4(i) = f_ed1(4);
end;

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

% check if all slots filled if v

if(VUS_voicetype(i)=='v')  % check  if v
	zeroindex = [];
	fcheck = [f1(i) f2(i) f3(i) ]; % dont care about f4
	zeroindex = find( fcheck== 0);
	allslotsfilled = isempty(zeroindex);
	if (VERB1 & allslotsfilled)
		disp('ALL SLOTS FILLED');
	elseif (VERB1 & ~allslotsfilled)
		s=sprintf('formant %d not filled', zeroindex);
		disp(s);
	end;
	passcnt = passcnt+1;
else
	allslotsfilled = 1;		
end;


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

end;	% end while loop

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

end;	% end main loop forward

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

% smooth out single zero values  - This is a mod by jmw

ct=0;
for i =2:m-1,
	if(f1(i)==0 & f1(i-1)~=0 & f1(i+1)~=0 )
		f1(i) = (f1(i-1) +f1(i+1) ) /2.0;
		ct = ct+1;
	end;

	if(f2(i)==0 & f2(i-1)~=0 & f2(i+1)~=0 )
		f2(i) = (f2(i-1) +f2(i+1) ) /2.0;
		ct = ct+1;
	end;

	if(f3(i)==0 & f3(i-1)~=0 & f3(i+1)~=0 )
		f3(i) = (f3(i-1) +f3(i+1) ) /2.0;
		ct = ct+1;
	end;
end;
s=sprintf('SMOOTHING: eliminate single zero data values ... %d events',ct);
disp(s);

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

% smooth out double zero values - This is a mod by jmw

ct = 0;
for i =3:m-1,
	if(f1(i)==0 & f1(i-1)==0 &  f1(i-2)~=0 & f1(i+1)~=0 )
		f1(i) = (f1(i-2) +f1(i+1) ) /2.0;
		f1(i-1) = f1(i);
		ct = ct+1;
	end;

	if(f2(i)==0 & f2(i-1)==0 & f2(i-2)~=0 & f2(i+1)~=0 )
		f2(i) = (f2(i-2) +f2(i+1) ) /2.0;
		f2(i-1) = f2(i);
		ct = ct+1;
	end;

	if(f3(i)==0 & f3(i-1)==0 & f3(i-2)~=0 & f3(i+1)~=0 )
		f3(i) = (f3(i-2) +f3(i+1) ) /2.0;
		f3(i-1) = f3(i);
		ct = ct+1;
	end;
end;
s=sprintf('SMOOTHING: eliminate double zero data values ... %d events',ct);
disp(s);

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

% find amplitudes 
disp('finding formant amplitudes ...');
N = 256;
a1 = zeros(m,1);
a2 = a1;
a3 = a1;
for i = 1:m,
	if(VUS_voicetype(i)=='v')  	
		% find formant amplitudes ...
		[h1,w1]=freqz([1],cofa(i,:),N);
		index1 = round(f1(i)/(5000/256));
		a1(i) = abs(h1(index1+1));
		index2 = round(f2(i)/(5000/256));
		a2(i) = abs(h1(index2+1));
		index3 = round(f3(i)/(5000/256));
		a3(i) = abs(h1(index3+1));
	end;
end;

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

% display results

titles = 1;

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

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

	subplot(313);
	stairs(range(:,1),[f1' f2' f3'  ]);
	axis([a(1) a(2) 0 5000]);
	grid on;
	if (titles)
		title('Formant Tracks: f1 f2 f3');
	end;

	subplot(312);
stairs(range(:,1),[20*log10(eps+a1) 20*log10(eps+a2) 20*log10(eps+a3)  ]);
	axis([a(1) a(2) 0 50]);
	grid on;
	if (titles)
		title('Amplitude Tracks: a1 a2 a3');
	end;

	drawnow;
end;

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

% save results to hard disk

s=sprintf('saving ./%s_Ftrk to disk ...', name);
disp(s);
s=sprintf('save temp/%s_Ftrk.mat f1 f2 f3 f4 a1 a2 a3', name);
eval(s);

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

% clean up time

clear MAXPASSES           f4                  n                   
clear N                   f_ed1                               
clear PLT                 f_initial           p                   
clear VERB1               f_nodup             passcnt             
clear VERB2               f_prev              power               
clear VUS_voicetype       f_prev_valid        power_index         
clear a                   f_prev_zero_index   r                   
clear a1                  fcheck              range               
clear a2                  file_string         residue             
clear a3                  h                   signal              
clear allslotsfilled      h1                  sort_power          
clear amp                 i                   target_frame        
clear cofa                index1              voicetype           
clear ct                  index2              w1                  
clear f1                  index3              zeroindex           
clear f2                  j                   
clear f3                  m  titles
                    
s=sprintf('clear %s',name);
eval(s);
clear s