Quick Start Guide

EasyClassifier Quick Start Guide

The audio classifier creation process is normally divided into three steps.

1. Get audios to use as examples.
2. Get(extract) caracteristics from these audio files.
3. Choose an algorithm that can categorize new audios based on what it learned from the examples.

Then you can use the classifier to classify new audios.

---

1. Categorize

In this part, you will add audios as examples. The categories are the labels of the audios.

Example: if I wanted to create a classifier that awnsers if an audio contains a whistle or not, I would have to create a category named "no" and another named "yes". In the yes category I would put the whistles and in the no category the counter examples.

-The more examples you have, the better the classifier will be. With a lot of different whistles, the learning algorithm is able to figure out which caracteristic is really important to determine whether the audio is a whistle or not.

2. Extract Features

In this part you can just click the extract automatically button.

it will extract some caracteristics of the audio

3.Train a classifier

In this part you have to choose which method you want to use to create a classifier. To test all and find the best given your examples, choose "Test all classifiers" and start.

A screen with a lot of text will show up. In the end you should look for the two lines marked on this picture.

These lines tell wich classifier is the best. You can now choose Train Classifier in the main screen.

The screen that wil pop asks for the name of the clasifier, the classifier type and the dimensionality reduction. The name may be anything you want, it will be used later to choose the classifier from a list. The next two arguments should be the same as the test all classifiers output.

Obs. Sometimes the dimensionality reduction does not work. If the output screen shows any errors try again with dimensionality reduction set to none.

---

Test classifier

Now you are ready to test your classifier. choose Test Classifier in this screen. Another main screen will be opened, it is intended to test a classifier.

to test the classifier, the first two steps are similiar.
1. Add the audios you want to clasify.
2. Extract the features pressing Extract automatically.
3. Choose a classifier from the list and press Classify.

The results window should pop and tell you what is the classification of the audio files you entered.