/// <summary>
/// Remove the silent segment from the given audio signal
/// </summary>
/// <param name="signal">original signal</param>
/// <returns>signal after removing the silent segment(s) from it</returns>
public static AudioSignal RemoveSilence(AudioSignal signal)
{
AudioSignal filteredSignal = new AudioSignal();
filteredSignal.sampleRate = signal.sampleRate;
filteredSignal.signalLengthInMilliSec = signal.signalLengthInMilliSec;
filteredSignal.data = MFCC.MFCC.RemoveSilence(signal.data, signal.sampleRate, signal.signalLengthInMilliSec, 20);
return filteredSignal;
}
/// <summary>
/// Open the given audio file and return an "AudioSignal" with the following info:
/// 1. data[]: array of audio samples
/// 2. sample rate
/// 3. signal length in milli sec
/// </summary>
/// <param name="filePath">audio file path</param>
/// <returns>AudioSignal containing its data, sample rate and length in ms</returns>
public static AudioSignal OpenAudioFile(string filePath)
{
WaveDecoder waveDecoder = new WaveDecoder(filePath);
AudioSignal signal = new AudioSignal();
signal.sampleRate = waveDecoder.SampleRate;
signal.signalLengthInMilliSec = waveDecoder.Duration;
Signal tempSignal = waveDecoder.Decode(waveDecoder.Frames);
signal.data = new double[waveDecoder.Frames];
tempSignal.CopyTo(signal.data);
return signal;
}
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;
}
static private AudioSignal openNISTWav(string filename)
{
int sample_rate = 0, sample_count =0, sample_n_bytes = 0;
StreamReader reader = new StreamReader(filename);
while(true)
{
string line = reader.ReadLine();
var splittedLine = line.Split(' ');
if (splittedLine[0] == "sample_count")
{
sample_count = int.Parse(splittedLine[2]);
}
else if (splittedLine[0] == "sample_rate")
{
sample_rate = int.Parse(splittedLine[2]);
}
else if (splittedLine[0] == "sample_n_bytes")
{
sample_n_bytes = int.Parse(splittedLine[2]);
}
else if (splittedLine[0] == "end_head")
break;
}
reader.Close();
byte[] wav = File.ReadAllBytes(filename);
//header offset.
int pos = 1024;
int samples = (wav.Length - pos) / sample_n_bytes; // 2 bytes per sample (16 bit sound mono)
int altsamples = sample_count / sample_n_bytes;
double[] data = new double[sample_count];
// Write to double array:
int i = 0;
while (pos < wav.Length)
{
data[i] = bytesToDouble(wav[pos], wav[pos + 1]);
pos += 2;
i++;
}
AudioSignal signal = new AudioSignal();
signal.sampleRate = sample_rate;
signal.data = data;
signal.signalLengthInMilliSec = (double) 1000.0 * sample_count / sample_rate ;
return signal;
}
/// <summary>
/// Extract MFCC coefficients of the sequence of frames for the given AudioSignal.
/// Each frame (feature) consists of 13 coefficients
/// </summary>
/// <param name="signal">Audio signal to extract its features</param>
/// <returns>Sequence of features (13 x NumOfFrames)</returns>
public static Sequence ExtractFeatures(ref AudioSignal signal)
{
//signal = RemoveSilence(signal);
return MFCC.MFCC.ExtractFeatures(signal.data, signal.sampleRate);
}
//Lower bounding function used for pruning
public static double LowerBound_Kim(AudioSignal signal, double firstElement, double lastElement, double minElement, double maxElement)
{
int size= signal.data.Length;
double maxElement1 = double.MinValue,
minElement1 = double.MaxValue,
firstElement1, lastElement1;
firstElement1 = signal.data[0];
lastElement1 = signal.data[size - 1];
for (int i = 0; i < size; i++)
{
maxElement1 = Math.Max(maxElement1, signal.data[i]);
minElement1 = Math.Min(minElement1, signal.data[i]);
}
double differenceBetweenFirsts = Math.Abs(firstElement1 - firstElement);
double differenceBetweenLasts = Math.Abs(lastElement1 - lastElement);
double differenceBetweenMinimums = Math.Abs(minElement1 - minElement);
double differenceBetweenMaximums = Math.Abs(maxElement1 - maxElement);
double lowerBoundValue = Math.Max(
Math.Max(differenceBetweenFirsts * differenceBetweenFirsts, differenceBetweenLasts * differenceBetweenLasts),
Math.Max(differenceBetweenMinimums * differenceBetweenMinimums, differenceBetweenMaximums * differenceBetweenMaximums));
//return maximum squared difference of the two sequences first, last, minimum and maximum elements
return lowerBoundValue;
}
public AddUser(Sequence sequence_, AudioSignal signal_)
{
sequence = sequence_;
signal = signal_;
InitializeComponent();
}