private void transform() {
for (int i = 1; i <= 10; i++) {
for (int j = 0; j < 10; j++)
for (int k = 1; k <= 10; k++) {
// strings.put(""+i+Utils.intToDigit(j)+k, levenshtein.Utils.transform(new
// RecordParser("records\\"+i+"\\"+Utils.intToDigit(j)+"\\"+Utils.intToDigit(j)+k+".txt").parse()));
strings.put(
"" + i + Utils.intToDigit(j) + k,
algorithm.levenshtein.Utils.transform(
new RecordParser(
"records/"
+ i
+ "/"
+ Utils.intToDigit(j)
+ "/"
+ Utils.intToDigit(j)
+ k
+ ".txt")
.parse()));
}
System.out.println(i + " transformed");
}
}
public void UI() {
for (int normalisation = 0; normalisation <= 3; normalisation++) {
tw = new TraceWriter("UI2DigitsStochasticLevenshtein3Norm" + normalisation + ".txt");
tw.println(
"-------------------------------------------------- User Dependent -------------------------------------------------");
tw.println();
for (int numberTrainingExamples = 3; numberTrainingExamples <= 3; numberTrainingExamples++) {
confusionMatrixes.clear();
for (int knn = 0; knn < numberTrainingExamples; knn++)
confusionMatrixes.add(new int[10][10]);
tw.println(
"------------------------------------------- numberTrainingExamples="
+ numberTrainingExamples
+ " ---------------------------------------");
tw.println();
userRecognitionRates = new double[10][9];
for (int user = 1; user <= 10; user++) {
UI(user, numberTrainingExamples, normalisation);
}
for (int knn = 0; knn < confusionMatrixes.size(); knn++) {
tw.println(
"normalisation="
+ normalisation
+ " numberTrainingExamples="
+ numberTrainingExamples
+ " knn="
+ (knn + 1)
+ " :");
tw.println();
tw.println(Utils.matrixToString(confusionMatrixes.get(knn)));
tw.println(Utils.matrixToStringForLatex(confusionMatrixes.get(knn)));
double[] informations = Utils.informations(confusionMatrixes.get(knn));
tw.println();
tw.println(
"normalisation="
+ normalisation
+ " numberTrainingExamples="
+ numberTrainingExamples
+ " knn="
+ (knn + 1)
+ " :");
tw.println();
for (int user = 0; user < 10; user++)
tw.println(
"Recognition rate for user" + (user + 1) + " = " + userRecognitionRates[user][knn]);
tw.println();
tw.println("Goodclass examples = " + informations[0]);
tw.println("Badclass examples = " + informations[1]);
tw.println("total examples = " + informations[2]);
tw.println("Recognition rate = " + informations[3]);
tw.println(
"----------------------------------------------------------------------------------------------------------------");
tw.println();
tw.println();
tw.println();
}
tw.println(
"================================================================================================================");
tw.println();
}
tw.close();
}
}
public void UI(int user, int numberTrainingExamples, int normalisation) {
List<int[][]> userConfusionMatrixes = new ArrayList<int[][]>();
for (int knn = 0; knn < numberTrainingExamples; knn++)
userConfusionMatrixes.add(new int[10][10]);
for (int firstUserIndex = 1; firstUserIndex <= 10; firstUserIndex++) {
String[] alphabet = {"0", "1", "2", "3", "4", "5", "6", "7"};
ConditionalRecognizer cr = new ConditionalRecognizer(alphabet);
for (int digitIndex = 0; digitIndex < 10; digitIndex++) {
int mod = 0;
int threshold = firstUserIndex + numberTrainingExamples;
for (int otherUser = firstUserIndex; mod * 10 + otherUser < threshold; otherUser++) {
if (otherUser == user) {
if (otherUser == 10) {
otherUser = 1;
mod = 1;
} else otherUser++;
threshold++;
}
cr.addTemplate(
Utils.intToDigit(digitIndex) + "",
strings.get("" + otherUser + Utils.intToDigit(digitIndex) + "1"));
if (otherUser == 10) {
otherUser = 0;
mod = 1;
}
}
}
cr.compile(normalisation);
for (int knn = numberTrainingExamples; knn <= numberTrainingExamples; knn++) {
for (int digitIndex = 0; digitIndex < 10; digitIndex++) {
for (int teIndex = 1; teIndex <= 10; teIndex++) {
String[] foundDigits =
cr.recognizeForAllKnn(
strings.get("" + user + Utils.intToDigit(digitIndex) + teIndex),
normalisation,
knn);
for (int i = 0; i < foundDigits.length; i++) {
String foundDigit = foundDigits[i];
confusionMatrixes.get(i)[digitIndex][Utils.digitToInt(foundDigit)]++;
userConfusionMatrixes.get(i)[digitIndex][Utils.digitToInt(foundDigit)]++;
}
System.out.println(
"norm="
+ normalisation
+ "|nte="
+ numberTrainingExamples
+ "|user="******"|fui="
+ firstUserIndex
+ "|char="
+ Utils.intToDigit(digitIndex)
+ " recognized");
}
}
}
}
for (int knn = 0; knn < numberTrainingExamples; knn++)
userRecognitionRates[user - 1][knn] = Utils.recognitionRate(userConfusionMatrixes.get(knn));
}