endpointdetect.m

bpsk 调制和解调 BPSK调制解调直接序列扩频系统源代码(白噪声及瑞利衰落误码率比较)

function [endPoint, volume, zcr, soundSegment] = endPointDetect(y, fs, nbits, plotOpt, epdParam) 
% endPointDetect: End point detection based on volume and zero-crossing rate 
%	Usage: [endPoint, volume, zcr, soundSegment] = endPointDetect(y, fs, nbits, plotOpt, epdParam) 
 
%	Roger Jang, 20041118 
 
if nargin<1, selfdemo; return; end 
if nargin<2, fs=8000; end 
if nargin<3, nbits=8; end 
if nargin<4, plotOpt=0; end 
if nargin<5, 
	epdParam.frameSize = round(fs/31.25);	% fs=8000 ===> frameSize=256; 
	epdParam.overlap = round(epdParam.frameSize/2); 
	epdParam.volRatio1=0.1; 
	epdParam.volRatio2=0.2; 
	epdParam.zcrShiftGain=4; 
	epdParam.zcrRatio=0.25; 
	epdParam.volRatio4zcr=0.05; 
end 
 
frameSize=epdParam.frameSize; 
overlap=epdParam.overlap; 
zcrRatio=epdParam.zcrRatio; 
 
if max(y)-min(y)<=2 
	error('The range of the given y vector is too small! Perhaps it should be converted to integer format...'); 
end 
 
% ====== Zero adjusted 
y = y-round(mean(y)); 
 
% ====== Take frames 
framedY  = buffer2(y, frameSize, overlap); 
frameNum = size(framedY, 2);			% Number of frames 
frameTime = frame2sampleIndex(1:frameNum, frameSize, overlap)/fs; 
 
% ====== Compute volume/zcr 
volume=frame2volume(framedY); 
[minVol, index]=min(volume); 
shiftAmount=epdParam.zcrShiftGain*max(abs(framedY(:,index)));	% 以最小音量之音框中的讯号最大绝对值的 epdParam.zcrShiftGain 倍为平移量来进行ZCR 
shiiftAmount=max(shiftAmount, 2);	 
zcr=frame2zcr(framedY, shiftAmount); 
 
% ====== Compute volume/zcr thresholds 
maxVol=max(volume); 
volTh2=(maxVol-minVol)*epdParam.volRatio2+minVol; 
%volTh1=(maxVol-minVol)*epdParam.volRatio1+minVol; 
%volTh1=minVol; 
zcrTh=max(zcr)*zcrRatio; 
zcrVolTh=(maxVol-minVol)*epdParam.volRatio4zcr+minVol; 
 
% ====== Identify voiced part that's larger than volTh2 
voicedIndex = volume>volTh2; 
soundSegment=findSegment(voicedIndex); 
 
%========求取噪音的平均值，然后作为声音的开始界限。 
%========同时确定有效声音的边界,从已知边界向两方扩展,直到frame的音量开始增加。 
NoiseIndex=volume<=volTh2; 
NoiseSegment=findSegment(NoiseIndex); 
j=0; 
NoiseTotal=0; 
for i=1:length(NoiseSegment) 
    NoiseTotal=sum(volume(NoiseSegment(i).begin:NoiseSegment(i).end)); 
    j=j+NoiseSegment(i).end-NoiseSegment(i).begin+1; 
end; 
volTh1=NoiseTotal/j;%声音噪音的平均值.此处还应该有个系数; 
 
 
     
 
% ====== Delete short sound clips 
%index = []; 
%for i=1:length(soundSegment), 
%	if (soundSegment(i).end-soundSegment(i).begin)<4 
%		index = [index, i]; 
%	end 
%end 
%soundSegment(index) = []; 
 
 
	% ====== Expansion 1: Expand end points to volume level1 (lower level) 
	for i=1:length(soundSegment), 
		head = soundSegment(i).begin; 
		while (head-1)>=1 & volume(head-1)< volume(head) &  volume(head-1)>volTh1, 
			head=head-1; 
		end 
		soundSegment(i).begin = head; 
		tail = soundSegment(i).end; 
		while (tail+1)<=length(volume) & volume(tail+1)>volTh1 & volume(tail+1)=1 & volume(head-1)>volTh1, 
% 			head=head-1; 
% 		end 
% 		soundSegment(i).begin = head; 
% 		tail = soundSegment(i).end; 
% 		while (tail+1)<=length(volume) & volume(tail+1)>volTh1, 
% 			tail=tail+1; 
% 		end 
% 		soundSegment(i).end = tail; 
% 	end 
% 	% ====== Expansion 2: Expand end points to include high zcr region 
% 	for i=1:length(soundSegment), 
% 		head = soundSegment(i).begin; 
% 		while (head-1)>=1 & zcr(head-1)>zcrTh & volume(head-1)>zcrVolTh			% Extend at beginning 
% 			head=head-1; 
% 		end 
% 		soundSegment(i).begin = head; 
% 		tail = soundSegment(i).end; 
% 		while (tail+1)<=length(zcr) & zcr(tail+1)>zcrTh & volume(tail+1)>zcrVolTh	% Extend at ending 
% 			tail=tail+1; 
% 		end 
% 		soundSegment(i).end = tail; 
% 	end 
 end 
 
% ====== 再跑一次 expansion 1, 以避免 expansion 1 & 2 的操作顺序不同而导致两个重叠的 segment 却有不同起点 or 终点 
 
% ====== Delete repeated sound segments 
index = []; 
for i=1:length(soundSegment)-1, 
	if soundSegment(i).begin==soundSegment(i+1).begin & soundSegment(i).end==soundSegment(i+1).end, 
		index=[index, i]; 
	end 
end 
soundSegment(index) = []; 
 
 
% ====== Transform sample-point-based index 
if length(soundSegment) ~=0, 
	for i=1:length(soundSegment), 
		soundSegment(i).beginSample = frame2sampleIndex(soundSegment(i).begin, frameSize, overlap); 
		soundSegment(i).endSample   = min(length(y), frame2sampleIndex(soundSegment(i).end, frameSize, overlap)); 
		soundSegment(i).beginFrame = soundSegment(i).begin; 
		soundSegment(i).endFrame = soundSegment(i).end; 
	end 
	endPoint=[soundSegment(1).beginSample, soundSegment(end).endSample];	% 取头尾 
else 
	endPoint = []; 
end 
%soundSegment=rmfield(soundSegment, 'begin'); 
%soundSegment=rmfield(soundSegment, 'end'); 
soundSegment=rmfield(soundSegment, 'duration'); 
 
% ====== 如果一个 segment 间隔不到0.3秒，而且过零率大于规定值，则删除，当作杂音。 
JianGe=[]; 
ZcrVol=[]; 
for i=1:length(soundSegment) 
    ZcrVol(i)=zcr(soundSegment(i).begin+round((soundSegment(i).end-soundSegment(i).begin)/2));%中间frame的过零率。 
end; 
for i=1:length(soundSegment) 
    JianGe(i)=(soundSegment(i).endSample-soundSegment(i).beginSample)/fs;%持续的时间。 
end; 
 
deletedIndex=[]; 
for i=1:length(JianGe) 
    if JianGe(i)<0.3 & ZcrVol(i)>zcrVolTh 
       deletedIndex=[deletedIndex,i]; 
    end; 
end; 
soundSegment(deletedIndex)=[]; 
 
 
% ====== 如果相邻的 segment 间隔不到0.2秒，则合并 
gap=[]; 
for i=1:length(soundSegment)-1 
	gap(i)=(soundSegment(i+1).beginSample-soundSegment(i).endSample)/fs; 
end 
group=findSegment(gap<0.1);%改为相邻0.1秒，不对的话再修改回来。 
deletedIndex=[]; 
for i=1:length(group) 
	soundSegment(group(i).begin).endSample=soundSegment(group(i).end+1).endSample; 
	soundSegment(group(i).begin).endFrame=soundSegment(group(i).end+1).endFrame; 
	deletedIndex=[deletedIndex, (group(i).begin+1):(group(i).end+1)]; 
end 
soundSegment(deletedIndex)=[]; 
 
%重新确定头尾。 
if length(soundSegment) ~=0, 
	endPoint=[soundSegment(1).beginSample, soundSegment(end).endSample];	% 取头尾 
else 
	endPoint = []; 
end 
 
if plotOpt, 
	axes1H=subplot(4,1,1); 
	time=(1:length(y))/fs; 
	plot(time, y); 
	axis([min(time), max(time), -2^nbits/2, 2^nbits/2]); 
	ylabel('Amplitude'); title('Waveform'); grid on 
	% Plot end points 
	yBound=[-2^nbits/2, 2^nbits/2]; 
	for i=1:length(soundSegment), 
		line(frame2sampleIndex(soundSegment(i).beginFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'm'); 
		line(frame2sampleIndex(  soundSegment(i).endFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'g'); 
	end 
 
	axes2H=subplot(4,1,2); 
	frameTime = frame2sampleIndex(1:frameNum, frameSize, overlap)/fs; 
	plot(frameTime, volume, '.-'); 
	line([min(frameTime), max(frameTime)], volTh1*[1 1], 'color', 'r'); 
	line([min(frameTime), max(frameTime)], volTh2*[1 1], 'color', 'r'); 
	line([min(frameTime), max(frameTime)], zcrVolTh*[1 1], 'color', 'r'); 
	axis tight 
	ylabel('Volume'); title('Volume'); grid on 
	% Plot end points 
	yBound = [min(volume) max(volume)]; 
	for i=1:length(soundSegment), 
		line(frame2sampleIndex(soundSegment(i).beginFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'm'); 
		line(frame2sampleIndex(  soundSegment(i).endFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'g'); 
	end 
 
	axes3H=subplot(4,1,3); 
	plot(frameTime, zcr, '.-'); 
	line([min(frameTime), max(frameTime)], zcrTh*[1 1], 'color', 'c'); 
	axis([min(frameTime), max(frameTime), 0, max(zcr)]); 
	ylabel('ZCR'); title('Zero crossing rate'); grid on 
	% Plot end points 
	yBound = [0 max(zcr)]; 
	for i=1:length(soundSegment), 
		line(frame2sampleIndex(soundSegment(i).beginFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'm'); 
		line(frame2sampleIndex(  soundSegment(i).endFrame, frameSize, overlap)/fs*[1,1], yBound, 'color', 'g'); 
	end 
 
	axes4H=subplot(4,1,4); 
	voicedIndex=endPoint(1):endPoint(2); 
	voicedTime=time(voicedIndex); 
	voicedY=y(voicedIndex); 
	voicedH=plot(voicedTime, voicedY); 
	axis([time(endPoint(1)), time(endPoint(2)), -2^nbits/2, 2^nbits/2]); 
	ylabel('Amplitude'); title('Voiced waveform'); grid on 
	 
	U.y=y; U.fs=fs; U.nbits=nbits; 
	U.axes1H=axes1H; U.axes2H=axes2H; U.axes3H=axes3H; U.axes4H=axes4H; 
	U.voicedIndex=voicedIndex; U.voicedH=voicedH; 
	U.voicedY=voicedY; U.voicedTime=voicedTime; 
	set(gcf, 'userData', U); 
	 
	uicontrol('string', 'Play all', 'callback', 'U=get(gcf, ''userData''); sound(U.y/(2^U.nbits/2), U.fs);'); 
	uicontrol('string', 'Play voiced', 'callback', 'U=get(gcf, ''userData''); sound(U.voicedY/(2^U.nbits/2), U.fs);', 'position', [100, 20, 60, 20]); 
 
	% Play the segmented sound 
%	head = soundSegment(1).beginFrame*(frameSize-overlap); 
%	tail = min(length(y), soundSegment(end).end*(frameSize-overlap)); 
%	thisY = y(head:tail); 
%	fprintf('His return to hear the cutted sound %g:', i); 
%	pause; 
%	fprintf('\n'); 
%	wavplay(thisY, fs, 'sync'); 
%	fprintf('\n'); 
end 
 
% ====== Self demo 
%function selfdemo 
%waveFile='清华大学资讯系.wav'; 
%plotOpt = 1; 
%[y, fs, nbits] = wavReadInt(waveFile); 
%endPoint = feval(mfilename, y, fs, nbits, plotOpt);