annotate Code/Descriptors/Matlab/Common/calculate_Silence.m @ 4:92ca03a8fa99 tip

Update to ICASSP 2013 benchmark

function [Limits] = calculate_Silence( x, fs, frameLength )

% Convert mono to stereo
if (size(x, 2)==2)
	x = mean(x')';
end
[m n]=size(x);
if(m>n)
    x=x';
end
% Window length and step (in seconds):
win = frameLength/fs;
step = win;

%%%%%%%%%%%%%%%%%%%%%%%%%%%
%  THRESHOLD ESTIMATION
%%%%%%%%%%%%%%%%%%%%%%%%%%%

% Weight = 10; % used in the threshold estimation method

% Compute short-time energy and spectral centroid of the signal:
Eor = ShortTimeEnergy(x, win*fs, step*fs);


% Apply median filtering in the feature sequences (twice), using 5 windows:
% (i.e., 250 mseconds)
E = medfilt1(Eor, 5); E = medfilt1(E, 5);

% normalise

E = E*(1/max(E));

% Get the average values of the smoothed feature sequences:
E_mean = mean(E);
% 
%Find energy threshold:
% [HistE, X_E] = hist(E, round(length(E) / 10));  % histogram computation
% [MaximaE, countMaximaE] = findMaxima(HistE, 3); % find the local maxima of the histogram


T_E = 0.0005; % determined empirically

% Thresholding:
Flags1 = (E>=T_E);

flags = Flags1;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%  SPEECH SEGMENTS DETECTION
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
count = 1;
WIN = 0;
Limits = [];
while (count < length(flags)) % while there are windows to be processed:
	% initilize:
	curX = [];	
	countTemp = 1;
	% while flags=1:
	while ((flags(count)==1) && (count < length(flags)))
		if (countTemp==1) % if this is the first of the current speech segment:
			Limit1 = round((count-WIN)*step*fs)+1; % set start limit:
			if (Limit1<1)	Limit1 = 1; end        
		end	
		count = count + 1; 		% increase overall counter
		countTemp = countTemp + 1;	% increase counter of the CURRENT speech segment
	end

	if (countTemp>1) % if at least one segment has been found in the current loop:
		Limit2 = round((count+WIN)*step*fs);			% set end counter
		if (Limit2>length(x))
            Limit2 = length(x);
        end
        
        Limits(end+1, 1) = Limit1;
        Limits(end,   2) = Limit2;
    end
	count = count + 1; % increase overall counter
end

%%%%%%%%%%%%%%%%%%%%%%%
% POST - PROCESS      %
%%%%%%%%%%%%%%%%%%%%%%%

% find the frame index of the start and end frames of non-silent segments
Limits = floor( Limits / frameLength );
pos = find(Limits == 0);
if(pos)
    Limits(pos) = 1;
end
    
% A. MERGE OVERLAPPING SEGMENTS:
RUN = 1;
while (RUN==1)
    RUN = 0;
    for (i=1:size(Limits,1)-1) % for each segment
        if (Limits(i,2)>=Limits(i+1,1))
            RUN = 1;
            Limits(i,2) = Limits(i+1,2);
            Limits(i+1,:) = [];
            break;
        end
    end
end

% B. Get final segments:
segments = {};
for (i=1:size(Limits,1))
    segments{end+1} = x(Limits(i,1)*frameLength:Limits(i,2)*frameLength); 
end