Tesis: Automatic method for voice recognition for the classification of vowels in the colombian sign language
In this repository an empirical comparison of different supervised learning algorithms for the automatic recognition of vocal sounds (a, e, i, o y u) is made to be used in favor of people with hearing disabilities, which make use of sign language Colombian. For this experiment, first a data set consisting of the pronunciation of the vowels is constructed; experiment with the characteristics obtained by LPC and MFCC; using SVM, neural networks and KNN, the results obtained with F1 score, precision, recall and conclusions are shown.
The methodological For this specific project is divided into 5 stages which will be presented below:
1) Collection of DataSet: In this part of the project the necessary data for the development of the project will be collected, which must contain specific characteristics; the file with the particular data and the name of said data.
https://drive.google.com/drive/folders/13cFXvZnebEP7pRXf0sIjxR_6hMRlFc93?usp=sharing
2) The extraction of characteristics, which by means of the code that as its name indicates it identifies the main or more relevant characteristics of the data for its later classification.
3) Afterwards, we will proceed to the sampling in which the data will be classified into two groups, the training data and the test data, assigning a percentage for each group and one that the allocation of these percentages can be somewhat subjective, normally these percentages oscillate 20% and 80%.
4) The fourth part of the methodological used for the development of the project is the classification of the data previously processed in the sub groups that are desired for the development of the project, which in this specific case are 5 different groups since they refer to the vowels.
5) To finish the last part of the methodology is the exit, where the result of the classification is transformed into an image that represents a vowel but in the Colombian sign language which can be easily identified by a person who is in condition Disability oral communication.