How to recognize gender by voice?

Determining a person’s gender as male or female, based upon a sample of their voice seems to be an easy task, because human can easily hear the difference between a male or female. However, designing a computer program to do this turns out to be a bit trickier.

Next, we explore statistical learning methods to recognize a person’s gender as male or female by voice. We build 4 models such as classification and regression tree, random forest, SVM and neural network to compare which one is better according to time consuming and area under ROC curve. In the end, considering both of accuracy and time consumption, we conclude that svm and neural network model for gender recognition are better.

Our dataset was built using thousands of samples of male and female voices which consist of 20 columns as features and 1 column labeled by their gender.

The following acoustic properties are our 20 features:

Next, we plot density pictures of each feature to learn which feature significantly differentiate gender.

After ploting density pictures of features, we conclude that these features such as sd, Q25, IQR, sp.ent, sfm and meanfun are significant.

For CART model, we get confusion matrix.

For random forest model, we get confusion matrix.

For SVM model, we get confusion matrix.

For neural network model, we get confusion matrix.

After building 4 models for recognition, we plot ROC curve for our models and present the area under ROC curve and time consumption of all algorithms in a table.

Considering both of accuracy and time consumption, we conclude that svm and neural network model for gender recognition are better.

Related posts:

A gente passa a vida inteira escrevendo historias, não sei todo mundo no sentido literal, como eu, mas certamente em algum outro sentido. É preciso uma imaginação praticamente nula pra não pensar no…

Functional programming is very useful it makes many development tasks easier to implement and makes our code more readable and concise and less prone to errors and bugs. This is only a big reason to…

As Smart Cities integram os setores público e privado, por meio das mais avançadas tecnologias, com o objetivo de tornar as cidades mais eficientes e seguras Uma cidade em que hospitais, delegacias…