%-------------------------------------------------------------------------------
% File Name   : epdone.m
% Description : This program detects the onset and offset of the speech for a single 
%               input wave file.
%-------------------------------------------------------------------------------              
% Author      : Prateek Bansal
%-------------------------------------------------------------------------------

clear all;
close all;

window_size = 10;    % Its in Milliseconds
end_silence = 100;   % Its in Milliseconds
back_off    = 250;   % In milliseconds 

% Create an output file
FILE = input(' Name of the Wave file to be processed(.wav):','s');

check = exist(FILE,'file');

if check ~= 2
   fprintf('!! The file is not present in the current directory ! \n');
   return;
end

% Read the Wave file
[input_sig Fs] = wavread(FILE);

% Normalized the input signal
input_sig = input_sig/max(input_sig);

% High Pass filtered at 100Hz to remove dc content and AC hum
W1 = 100 *(2/Fs) ; 

% Low Pass Filtered at 4KHz
W2 = (4*10^3)*(2/Fs);

% Design the bandpass filter
[B,A]= butter(10,[W1,W2]);

% Filtered Signal Output
filt_inpsig = input_sig;

% Normalize the filtered signal
filt_inpsig = filt_inpsig/max(filt_inpsig);

% Length of filtered data
signal_length = length(filt_inpsig);

% Define the time axis
time_t = 0 : 1 : (signal_length-1); 

% WindowSize in Samples
win_samples = round((Fs*window_size)/1000);

% Define the Hamming Window
ham_window = hamming(win_samples);

j = 1;
for i = 1: win_samples : signal_length - win_samples
   temp = filt_inpsig(i:i+win_samples-1);
   temp2 = temp.*ham_window;
   temp2 = abs(filter(B,A,temp2));
   energy(j) = sum(temp2);
   j = j+1;
end

% Compute of End of Silence in Samples
end_silence_sample = round((end_silence*Fs)/1000);

% Define the Silence Range
silenceRange = 1:length(1:win_samples:end_silence_sample - win_samples);

% Energy Thresholds
% IMN (silence energy) is average energy for initial 100ms of signal.

%      IMX=max(energy);                       % peak Energy
%      IMN=mean(energy(silenceRange));        % Silence energy
%      I1=0.03*(IMX-IMN)+IMN;                 

%      I2=4*IMN;
%      ITL=min(I1,I2);                        % Lower Threshold
ITL = mean(energy) * 0.2;
ITU=5*ITL;                             % Upper Threshold

% Determination of End Points

N1=0;        % Start point initial estimate
N2=0;        % End point initial estimate

duration=length(energy);

done=0;

% Estimation of the start point based on energy considerations

for m=1:duration
   if and(energy(m)>=ITL,~done)
      for i=m+2:duration
         if energy(i)<ITL
            break
         else   
            if energy(i)>=ITU
               if ~done
                  N1=i-(i==m);
                  done=1;
               end
               break
            end
         end
      end
   end   
end

done=0;

% Estimation of the end point based on energy considerations

for m=duration:-1:1
   if and(energy(m)>=ITL,~done)
      for i=m-2:-1:1
         if energy(i)<ITL
            break;
         else
            if energy(i)>=ITU
               if ~done
                  N2=i+(i==m);
                  done=1;
               end
               break
            end
         end
      end   
   end
end

warpRatio=round(length(filt_inpsig)/length(energy));
N1_w=N1*warpRatio;
N2_w=N2*warpRatio;

% Plotting the signal
p = 1;

if p
   subplot(2,1,1);
   plot(time_t,input_sig);
   hold on;
   line([N1_w,N1_w],[-1,1],'Color',[1 0 1]);
   line([N2_w,N2_w],[-1,1],'Color',[1 0 1]);
   title(' End Points of the Speech Signal');
   subplot(2,1,2);
   plot(1:length(energy),energy);
   title(' Energy of the signal');
end

% Evaluate Sppech_on and Speech_off
Speech_on = round((N1_w*1000)/Fs);
Speech_off = round((N2_w*1000)/Fs);
Speech_on  = round(Speech_on - 0.05*Speech_on);

fprintf(1,'\n Speech Onset for %s  file : %d msec \n', FILE,Speech_on );
fprintf(1,' Speech offset for %s  file : %d msec\n', FILE,Speech_off);