LOGIC.txt

(1)
Clipped Sir provided voice data to individual 8 parts.
and worked on that.

take necessary file pointers for input, output, normalised input, voice extraction.
-----------

(2)
Read from file provided - all the samples data. 

DC Shift Steps:
read each samples. also can be saved in array samples[60000]
SUM over all x_i (+ve or -ve) and then taking
dcShift = sum_above/totalSamples
(total samples after ignorning header data if there or other)

meanwhile in same loop calculate max AMP.

Normalization:
(eachX_i - dcShift)*5000/ (max of abs(AMP)))
simultaneously output it in norm_file
or save in norm_samples_array[60000] (to be able to use easy without reading file)

(3)
First we have to loop over normalised data
calculating 

Energy_i = x_i * x_i (over 100 samples for one frame i )
AvgEnergy_i = Energy_i/100(Frames)

cntZCR_i = (total X_i samples crossing 0 for 100 frames)
{ cntZCR[i] +=  (s_i_1*s_i < 0);
  s_i_1 = s_i;
}
AvgZCR_i = cntZCR_i / 100

TotEnergy = Sum of all AvgEnergy_i for all frames

then
avgEnergy[i]/=sizeFrame;
avgZCR[i] = cntZCR[i]/sizeFrame;

//noise
Calculating ZCR and Energy of First Few (5-10) Frames (Noise) (after 2nd step)

AvgNoise_ZCR = SUM_avgZCR_first_few / first_few_frames

countNoise_ZCR = SUM_countZCR_first_few / first_few_frames
AvgNoise_Energy = SUM_NoiseE_first_few / first_few_frames

ZCR_Threshold = countNoise_ZCR/first_few_frames
Energy_Threshold = AvgNoise_Energy * 3 ((3-10%))

(4)
//one logic 
//if it works, you can from start or end marker, segregate data and run in cool edit to check its accuracy

bool flag =false;
Loop over all the frames(i) (skip last 2-3 frames)
   IF [ AvgEnergy_i+1 > Energy_Threshold OR AvgZCR_i+1 < ZCR_Threshold ]
	IF [ Check AvgEnergy of next i+2 and i+3 also > Energy_Threshold && flag=false]
		(increasing curve)
		startFrameMarker(seg) = i (frame num) (multiple segment if any)
		flag =  trur start detected
   ELSEIF [ 
		flag detected &&  AvgZCR_i > ZCR_Threshold  && AvgEnergy_i < Energy_Threshold
		&& AvgEnergy_i-1 < Energy_Threshold && AvgEnergy_i-2 < Energy_Threshold
	  ] (decreasing curve)

		stopFrameMarker(seg++) = i
		flag = false (to detect other marker)
// another logic
for(int i=0; i<cntTotFrames-3; ++i){
	if(!flag && avgEnergy[i+1] > thresholdEnergy && avgEnergy[i+2] > thresholdEnergy && avgEnergy[i+3] > thresholdEnergy){
					start = i;
					flag = 1;
	}
	else if(flag && avgEnergy[i+1] <= thresholdEnergy && avgEnergy[i+2] <= thresholdEnergy&& avgEnergy[i+3] <= thresholdEnergy){
		end = i;
		flag = 0;
		break;
	}
}
if(flag == 1) end = cntTotFrames - 5; //if end is not found then making the last frame - 3 as the end marker for the word
	long startSample= (start+1)*sizeFrame;
	long endSample= (end+1)*sizeFrame;
	long totFramesVoice = end-start+1;


(5)Also using Markers segmented data can be put into their separate file
(6) Opening New file and putting data in an array OR 
putting data in array from the point of normalization 
I think both will work, 2nd one being easy I think, have to maintain single array for one original file.

(7-8)Then for each segment from startFrameMarker to stopFrameMarker
AvgEnergy and AvgZCR can be calculated by looping around 
calculating first total things from start to stop and then dividing by total frames between them
	
float voiceAvgEnergy=0, voicecntZCR=0, voiceavgZCR=0;
	for(i=start;i <= end;i++){
			voiceAvgEnergy+=avgEnergy[i];
			voicecntZCR+=cntZCR[i];
			voiceavgZCR+=avgZCR[i];
	}
	voiceAvgEnergy/=totFramesVoice;
	voicecntZCR/=totFramesVoice;
	voiceavgZCR/=totFramesVoice;


(9)
//logic 1
Last 40% frames has high CNT of ZCR then it is YES ELSE NO
Last40%AvgZCR = Last40%SUMofZCR/40%frames
if(avgLastFourtyPercZCR >= lastFourtyPercThreshold(30,40 anything that works for you))
		printf("  YES signal\n"); 
	else
		printf("  NO signal\n");
//logic 2
count fricatives.
 for ( i=start; i<=end-1; ++i){
		if(cntZCR[i] > 15 && cntZCR[i+1] > 15)
			cntFricatives++;
	}
if(cntFricatives >= totFramesVoice * 0.3)	
		printf("= YES signal\n"); 
	else
		printf("  NO signal\n"); 
//logic 3

		for(i=end; i >= start;i--)
		{
			if(avgZCR[i] > voiceavgZCR && avgEnergy[i] < voiceAvgEnergy)
				rHighZCRcount++;
			if(avgZCR[start+end-i] < voiceavgZCR && avgEnergy[i] < voiceAvgEnergy)
				lLowZCRcount++;
		}

		midframe=(end+start)/2;
		if(midframe == 0)
			printf("Non speech signal\n");
		else if(rHighZCRcount > 10*300/sizeFrame)
			printf(" Voice Data is YES signal\n");	
		else
			printf(" Voice Data is NO signal\n");
		printf("rHighZCRcount(%ld) > %d\n", rHighZCRcount,10*300/sizeFrame );
10
close file streams
print necessary data whenever necessary