//DTW without pruning
public static double DTW_Distance(Sequence sequence1, Sequence sequence2)
{
int numberOfFrames_Sequence1 = sequence1.Frames.Count();
int numberOfFrames_Sequence2 = sequence2.Frames.Count();
double[,] DTW = new double[3, numberOfFrames_Sequence2 + 1];
for (int i = 0; i <= numberOfFrames_Sequence2; i++)
{
DTW[0, i] = DTW[1, i] = DTW[2, i] = double.MaxValue;
}
DTW[0, 0] = 0;
//Applying dimension compression to DTW array
for (int i = 2; i <= numberOfFrames_Sequence1 + 1; i++)
{
for (int j = 1; j <= numberOfFrames_Sequence2; j++)
{
double cost = distance(sequence1.Frames[i - 2], sequence2.Frames[j - 1]);
DTW[i % 3, j] = cost + Math.Min(DTW[i % 3, j - 1], //horizontal, stretching
Math.Min(DTW[(i + 1) % 3, j - 1], //diagnoal, aligning
DTW[(i + 2) % 3, j - 1])); //far diagonal, shrinking
}
}
return DTW[(numberOfFrames_Sequence1 + 1) % 3, numberOfFrames_Sequence2];
}
public MainForm()
{
InitializeComponent();
// Configure the wavechart
chart.SimpleMode = true;
chart.AddWaveform("wave", Color.Green, 1, false);
updateButtons();
List<User> listOfUsers = TestcaseLoader.LoadTestcase1Testing("E:\\Complete SpeakerID Dataset\\TestingList.txt");
foreach (User user in listOfUsers)
{
int counter = 0, counter1 = 0;
foreach (AudioSignal signal in user.UserTemplates)
{
AudioSignal signal_ = signal;
sequence = AudioOperations.ExtractFeatures(ref signal_);
ClosestMatch match = FileOperations.GetUserName(sequence, signal, true);
ClosestMatch match1 = FileOperations.GetUserName(sequence, signal, false);
Console.WriteLine(user.UserName + " : " + match.Username + ", " + match1.Username);
if (user.UserName == match.Username) counter++;
if (user.UserName == match1.Username) counter1++;
}
Console.WriteLine("====== " + counter + " out of " + user.UserTemplates.Count + " with pruning.");
Console.WriteLine("====== " + counter1 + " out of " + user.UserTemplates.Count + " without pruning.");
}
Console.WriteLine("Done");
}
public static Sequence ExtractFeatures(double[] pSignal, int samplingRate)
{
Sequence sequence = new Sequence();
double[,] mfcc = MATLABMFCCfunction(pSignal, samplingRate);
int numOfFrames = mfcc.GetLength(1);
int numOfCoefficients = mfcc.GetLength(0);
Debug.Assert(numOfCoefficients == 13);
sequence.Frames = new MFCCFrame[numOfFrames];
for (int i = 0; i < numOfFrames; i++)
{
sequence.Frames[i] = new MFCCFrame();
for (int j = 0; j < numOfCoefficients; j++)
{
sequence.Frames[i].Features[j] = mfcc[j, i];
}
}
return sequence;
}
public static void SaveSequenceInDatabase(Sequence toBeSavedSequence, string username, AudioSignal signal)
{
//UPDATE
//you should save the four values in the last row before the username, with the order (first, last, min, max) respectively
FileStream SavingStream = new FileStream("savedSequences.txt", FileMode.Append);
StreamWriter Saving = new StreamWriter(SavingStream);
double TempFeature = 0f;
StringBuilder FramesRow = new StringBuilder();
int size = signal.data.Length;
double maxElement = double.MinValue,
minElement = double.MaxValue,
firstElement, lastElement;
for (int i = 0; i < 13; i++)
{
for (int j = 0; j < toBeSavedSequence.Frames.Length; j++)
{
TempFeature = toBeSavedSequence.Frames[j].Features[i];
FramesRow.Append(TempFeature.ToString() + "|");
}
Saving.WriteLine(FramesRow);
FramesRow.Clear(); //clear it to start a new row (VIP)
}
firstElement = signal.data[0];
lastElement = signal.data[size - 1];
for (int i = 0; i < size; i++)
{
maxElement = Math.Max(maxElement, signal.data[i]);
minElement = Math.Min(minElement, signal.data[i]);
}
// UPDATE: On 26/12 @ 6:50 - Writing the 4 values into the file
Saving.WriteLine(firstElement + " " + lastElement + " " + minElement + " " + maxElement);
Saving.WriteLine("Username:" + username);
Saving.Close();
}
//Identify button opens the file explorer to choose a pre existing audio file or recorded sound to be identified
private void btnIdentify_Click(object sender, EventArgs e)
{
ClosestMatch User = new ClosestMatch();
if (sequence != null && RecordRadio.Checked == false)
{
var watch = Stopwatch.StartNew();
User = FileOperations.GetUserName(sequence, signal, WithPruningRadioBTN.Checked);
watch.Stop();
var elapsedMs = watch.ElapsedMilliseconds;
Console.WriteLine("Elapsed Milliseconds = " + elapsedMs);
MessageBox.Show("Username: "******"\nWith Minimum Difference: " + User.MinimumDistance.ToString());
}
else if (SavedRadio.Checked)
{
OpenFileDialog open = new OpenFileDialog();
if (open.ShowDialog() == DialogResult.OK)
{
isRecorded = false;
path = open.FileName;
//Open the selected audio file
signal = AudioOperations.OpenAudioFile(path);
sequence = AudioOperations.ExtractFeatures(ref signal);
var watch = Stopwatch.StartNew();
User = FileOperations.GetUserName(sequence, signal, WithPruningRadioBTN.Checked);
watch.Stop();
var elapsedMs = watch.ElapsedMilliseconds;
Console.WriteLine("Elapsed Milliseconds = " + elapsedMs);
MessageBox.Show("Username: "******"\nWith Minimum Difference: " + User.MinimumDistance.ToString());
}
}
//Dev: Omar Moataz Abdel-Wahed Attia
else
{
if (isRecorded)
{
InitializeDecoder(); //Initializes a decoder to get the value of the recorded stream.
AudioSignal signal = new AudioSignal(); //Signal sent to Feature extraction function.
signal.data = new double[this.decoder.frames]; //Reserve space for double array that will be filled later.
signal.sampleRate = this.decoder.GetTempSignal().SampleRate;
//TempSignal has the double array I need to extract features, Check function Decoder::getWholeSignal() for more explanation.
this.decoder.GetTempSignal().CopyTo(signal.data);
//Copies the values of the signal to an object "signal" of type AudioSignal which is sent to feature extraction.
sequence = AudioOperations.ExtractFeatures(ref signal);
//Get name of user that has the closest match.
var watch = Stopwatch.StartNew();
User = FileOperations.GetUserName(sequence, signal, WithPruningRadioBTN.Checked);
watch.Stop();
var elapsedMs = watch.ElapsedMilliseconds;
Console.WriteLine("Elapsed Milliseconds = " + elapsedMs);
MessageBox.Show("Username: "******"\nWith Minimum Difference: " + User.MinimumDistance.ToString());
}
else
{
MessageBox.Show("Please record your voice first!"); //In case the user tries to identify without recording any sound.
}
}
sequence = null;
updateButtons();
}
public void openToolStripMenuItem_Click(object sender, EventArgs e)
{
OpenFileDialog open = new OpenFileDialog();
if (open.ShowDialog() == DialogResult.OK)
{
isRecorded = false;
path = open.FileName;
//Open the selected audio file
signal = AudioOperations.OpenAudioFile(path);
sequence = AudioOperations.ExtractFeatures(ref signal);
updateButtons();
}
}
private void saveFileDialog1_FileOk(object sender, System.ComponentModel.CancelEventArgs e)
{
if (this.encoder != null)
{
Stream fileStream = saveFileDialog1.OpenFile();
this.encoder.Save(fileStream);
path = saveFileDialog1.FileName;
signal = AudioOperations.OpenAudioFile(path);
sequence = AudioOperations.ExtractFeatures(ref signal);
}
}
//======================================================
/*
Dev: Omar Moataz Abdel-Wahed Attia
Last Edit: 12/8/2015
To understand the code in function GetUserName, you need to understand the file structure
I'm looping over.
The file will contain 13 lines which represent a Frame (0-12) (Each Column is a frame)
on the 14th line, it will contain the name of the person that's tied to the previous sequence.
*/
//======================================================
public static ClosestMatch GetUserName(Sequence sequence, AudioSignal signal, bool pruned)
{
ClosestMatch User = new ClosestMatch();
//Opening the file.
using (StreamReader Reader = new StreamReader("savedSequences.txt"))
{
//Initializing a new sequence.
Sequence ToBeCompared = new Sequence();
//This line string contains every line I go through in the file
string Line;
//Holds the value of the current frame
int Index = 0;
bool flag = true, Updated = false;
//Variables used in lowerbounding
double FirstElement = 0, LastElement = 0, MaxElement = 0, MinElement = 0;
while ((Line = Reader.ReadLine()) != null)
{
if (Index == 13)
{
double TrueDistance;
string[] Temp = Line.Split(' '); /*Just a string array that holds the values I'll take into FirstElement,
LastElement, MinElement and MaxElement.*/
FirstElement = double.Parse(Temp[0]);
LastElement = double.Parse(Temp[1]);
MinElement = double.Parse(Temp[2]);
MaxElement = double.Parse(Temp[3]);
if (pruned)
{
double LowerBoundDistance = DynamicTimeWarpingOperations.LowerBound_Kim(signal, FirstElement, LastElement, MinElement, MaxElement);
if (LowerBoundDistance > User.MinimumDistance) goto skip;
TrueDistance = DynamicTimeWarpingOperations.Pruned_DTW_Distance(ToBeCompared, sequence);
}
else
{
TrueDistance = DynamicTimeWarpingOperations.DTW_Distance(ToBeCompared, sequence);
}
//Here I compare the two Distances together to see if I need to update the minimum or not.
if (TrueDistance < User.MinimumDistance)
{
//Here I update the minimum distance between two values.
User.MinimumDistance = TrueDistance;
Updated = true;
}
skip:
//This is a reinitialization just to clear out old values from the previous iteration
flag = true;
ToBeCompared = new Sequence();
}
else if(Index == 14)
{
if (Updated)
{
//I update the name of the person to line because on the 13th index line, it'll have the name of the person.
User.Username = Line.Substring(9, Line.Length - 9);
}
Updated = false; //resetting the update value.
Index = -1;
//So, it goes back to 0 when the loop continues.
}
else
{
string[] ExtractedStringsFromLine = Line.Split('|');
if (flag == true)
{
ToBeCompared.Frames = new MFCCFrame[ExtractedStringsFromLine.Length - 1];
}
//Here I split all the values from every line to an array of Strings.
for (int i = 0; i < ExtractedStringsFromLine.Length - 1; i++)
{
if (flag == true)
{
ToBeCompared.Frames[i] = new MFCCFrame();
}
ToBeCompared.Frames[i].Features[Index] = double.Parse(ExtractedStringsFromLine[i]);
}
flag = false;
}
//I increment the index of the 2D array for the next iteration through the file.
++Index;
}
}
//I return type ClosestMatch.
return User;
}
public AddUser(Sequence sequence_, AudioSignal signal_)
{
sequence = sequence_;
signal = signal_;
InitializeComponent();
}