static void Test(double[][] testData, string pathModel)
{
Console.WriteLine("Montando treino e target");
SVMAudio mySVM = new SVMAudio();
mySVM.LoadSVMFromFile(pathModel);
int qtdAudio = testData.Length;
int featuresSize = testData[0].Length - 1;//menos a classe
Image <Gray, float> features = new Image <Gray, float>(featuresSize, qtdAudio);
Image <Gray, int> target = new Image <Gray, int>(1, qtdAudio);
Image <Gray, float> results = new Image <Gray, float>(1, qtdAudio);
for (int audioIndex = 0; audioIndex < testData.Length; audioIndex++)
{
for (int j = 0; j < featuresSize; j++)
{
features.Data[audioIndex, j, 0] = (float)testData[audioIndex][j];
}
target.Data[audioIndex, 0, 0] = (int)testData[audioIndex][featuresSize]; // posição da classe
}
Console.WriteLine("Testando");
mySVM.Predict(features, results);
int cont = 0;
int[] qtd = new int[11];
for (int i = 0; i < qtd.Length; i++)
{
qtd[i] = 0;
}
for (int i = 0; i < qtdAudio; i++)
{
Console.WriteLine("Correct: " + target.Data[i, 0, 0] + " Predicted: " + results.Data[i, 0, 0]);
if (target.Data[i, 0, 0] == (int)results.Data[i, 0, 0])
{
cont++;
}
qtd[target.Data[i, 0, 0]]++;
}
Console.WriteLine("Acc: " + (float)cont / qtdAudio);
for (int i = 0; i < qtd.Length; i++)
{
Console.WriteLine("Quantidade " + i + ": " + qtd[i]);
}
Console.ReadKey();
}
static void Train(double[][] trainData, string pathModel)
{
Console.WriteLine("Montando treino e target");
SVMAudio mySVM = new SVMAudio();
int qtdAudio = trainData.Length;
int featuresSize = trainData[0].Length - 1;//menos a classe
Image <Gray, float> features = new Image <Gray, float>(featuresSize, qtdAudio);
Image <Gray, int> target = new Image <Gray, int>(1, qtdAudio);
for (int audioIndex = 0; audioIndex < trainData.Length; audioIndex++)
{
for (int j = 0; j < featuresSize; j++)
{
features.Data[audioIndex, j, 0] = (float)trainData[audioIndex][j];
}
target.Data[audioIndex, 0, 0] = (int)trainData[audioIndex][featuresSize]; // posição da classe
}
Console.WriteLine("Treinando");
mySVM.Train(features, target);
mySVM.SaveSVMToFile(pathModel);
}
static float RecognizeAudio(string path, string svmModel)
{
SVMAudio mySVM = new SVMAudio();
mySVM.LoadSVMFromFile(svmModel);
double[] y = AudioFeatures.load2(path);
double[,] mfcc_mass = MFCC_calculating.MFCC_20_calculation(y);
double[] features_vector = FeaturesFromMFCC(mfcc_mass);
int features_size = 100;
Image <Gray, float> features = new Image <Gray, float>(features_size, 1);
Image <Gray, float> results = new Image <Gray, float>(1, 1);
AudioFeatures.NormalizeFromFile(@"C:\Projeto\autoleitura\paulorenato\AudioCSharp\Modelo\scaleSVMTrain.txt", features_vector);
for (int i = 0; i < features_size; i++)
{
features.Data[0, i, 0] = (float)features_vector[i];
}
mySVM.Predict(features, results);
return(results.Data[0, 0, 0]);
}
