www.pudn.com > audioProcessingtoolbox.rar > epdByVolHod.m

function [epInSampleIndex, epInFrameIndex, segment, zeroOneVec, volume, hod, vh] = epdByVolHod(y, fs, nbits, epdParam, plotOpt) 
% epdByVolHod: EPD based on volume and HOD (high-order difference) 
%	Usage: [epInSampleIndex, epInFrameIndex, segment, zeroOneVec, volume, hod, vh] = epdByVolHod(y, fs, nbits, epdParam, plotOpt) 
%		epInSampleIndex: two-element end-points in sample index 
%		epInFrameIndex: two-element end-points in frame index 
%		segment: segment of voice activity 
%		y: input audio signals 
%		fs: sampling rate 
%		nbits: no. of bits 
%		epdParam: parameters for EPD 
%		plotOpt: 0 for silence operation, 1 for plotting 
% 
%	Example: 
%		waveFile='SingaporeIsAFinePlace.wav'; 
%		[y, fs, nbits] = wavReadInt(waveFile); 
%		epdParam=epdParamSet(fs); 
%		plotOpt = 1; 
%		[epInSampleIndex, epInFrameIndex, segment] = epdByVolHod(y, fs, nbits, epdParam, plotOpt); 
 
%	Roger Jang, 20070323 
 
if nargin<1, selfdemo; return; end 
if nargin<2, fs=16000; end 
if nargin<3, nbits=16; end 
if nargin<4 | isempty(epdParam), epdParam=epdParamSet(fs); end 
if nargin<5, plotOpt=0; end 
 
if size(y, 2)~=1, error('y is not mono!'); end 
 
frameSize=epdParam.frameSize; 
overlap=epdParam.overlap; 
minSegment=round(epdParam.minSegment*fs/(frameSize-overlap)); 
maxSilBetweenWord=round(epdParam.maxSilBetweenWord*fs/(frameSize-overlap)); 
%minLastWordDuration=round(epdParam.minLastWordDuration*fs/(frameSize-overlap)); 
 
% ====== Compute volume/hod 
frameMat=buffer2(y, frameSize, overlap);	% frame blocking 
frameMat=frameZeroMean(frameMat, 2); 
frameNum=size(frameMat, 2);			% no. of frames 
volume=frame2volume(frameMat, 1);		% method=1 
hod=frame2ashod(frameMat, epdParam.diffOrder); 
 
% ====== Compute vh thresholds 
vh=volume*epdParam.volWeight+(1-epdParam.volWeight)*hod; 
temp=sort(vh); 
index=round(frameNum/32); if index==0, index=1; end 
vhMin=temp(index); 
vhMax=temp(frameNum-index+1);			% To avoid qiYin 
vhTh=(vhMax-vhMin)/epdParam.vhRatio+vhMin; 
%fprintf('vhMin=%g, vhMax=%g, vhTh=%g\n', vhMin, vhMax, vhTh); 
 
% ====== Identify voiced part that's larger than volTh2 
segment=findSegment(vh>vhTh); 
 
% ====== Delete short sound clips 
index = []; 
for i=1:length(segment), 
	if segment(i).duration<=minSegment 
		index = [index, i]; 
	end 
end 
segment(index) = []; 
 
% ====== If the sil between the last two segment is too big, delete the last segment 
if length(segment)>1 
	if segment(end).begin-segment(end-1).end>maxSilBetweenWord 
%		if segment(end).duration<=minLastWordDuration		% Not used! 
			segment(end)=[]; 
%		end 
	end 
end 
 
% ====== Check if segment 1 is noise 
%if length(segment)>=2 
%	sil=segment(2).begin-segment(1).end-1; 
%	if (sil>epdParam.minSil) && (2*segment(1).duration=2 
%	sil=segment(end).begin-segment(end-1).end-1; 
%	if (sil>epdParam.minSil) && (segment(end-1).duration>2*segment(end).duration) 
%		segment(end)=[]; 
%	end 
%end 
 
 
% Use segment2 for further processing. Keep segment for plotting 
segment2=segment; 
% If a sil is longer than its neighboring segment, delete the segment 
while 0 
	silDuration=[]; 
	for i=1:length(segment2)-1 
		silDuration(i)=segment2(i+1).begin-segment2(i).end-1; 
	end 
	if isempty(silDuration), break; end 
	[maxSil, index]=max(silDuration); 
%	if maxSil>epdParam.minSil	% max silence too long ===> delete one of its neighboring segment 
	if maxSil>min(segment2(index).duration, segment2(index+1).duration)		% max silence too long ===> delete one of its neighboring segment 
		if segment2(index).duration1, U.y=U.y/(2^nbits/2); end 
	if ~isempty(epInSampleIndex) 
		U.voicedY=U.y(epInSampleIndex(1):epInSampleIndex(end)); 
	else 
		U.voicedY=[]; 
	end 
	set(gcf, 'userData', U); 
	uicontrol('string', 'Play all', 'callback', 'U=get(gcf, ''userData''); sound(U.y, U.fs);'); 
	uicontrol('string', 'Play voiced', 'callback', 'U=get(gcf, ''userData''); sound(U.voicedY, U.fs);', 'position', [100, 20, 60, 20]); 
end 
 
% ====== Self demo 
function selfdemo 
waveFile='SingaporeIsAFinePlace.wav'; 
[y, fs, nbits] = wavReadInt(waveFile); 
epdParam=epdParamSet(fs); 
plotOpt = 1; 
[epInSampleIndex, epInFrameIndex, volume, hod, vh] = epdByVolHod(y, fs, nbits, epdParam, plotOpt);