This document presents research on using machine learning algorithms to classify gender based on voice characteristics. It analyzed a dataset of 3,168 voice samples labeled as male or female using acoustic features. Three classifiers - decision tree, random forest, and logistic regression - were tested on 80% of the data and evaluated on the remaining 20%. Random forest achieved the highest accuracy of 98.5%, while decision tree and logistic regression achieved 97.3% and 92.7% accuracy respectively. The document concludes random forest performed best on this dataset for gender classification based on voice.

1. Al-Nahrain University
College of Information Engineering and Communications
Department of Information Engineering and Communications
Gender Detection Based on Voice and Speech Recognition Using Different
Machine LearningAlgorithms
By:
Mohammed Wahhab Abdulrazzaq

2. Introduction
Each vocal sound has physical characteristics that can be related
to perceptual evaluations of that sound.
Frequency (measured in Hertz, Hz), sound level (measured in
decibels, dB), acoustic characteristics.
Here we will be focused only on the measurement of physical
characteristics of the sound (Acoustic)

3. Natural Sound Generation

4. Dataset
In order to analyze gender by voice and speech, a training
database is required.
A database was built using 3168 of samples of male and
female voices, each labeled by their gender of male or
female.
This Dataset contains 3168 rows and 21 columns (20
columns for each feature and one label column for the
classification of male or female).

5. Dataset Cont..
➢The voice physical characteristics (Acoustic) are obtained
using Specan function using Warble package in R
language.
➢Meanfreq, Sd, Median, Q25, Q75, IQR, Skew, kurt,
sp.ent, sfm, Mode centroid, meanfun, minfun, maxfun,
meandom, mindom, minidom, maxdom, dfrange,
modindx.

6. Machine Learning Classification
Decision Tree Classification
supervised learning method used
for classification and regression.
The goal is to create a model that predicts the value of a target
variable by learning simple decision rules inferred from the
data features.
Decision tree algorithms: ID3 , C4.5 , C5.0, CART

7. Machine Learning Classification Cont..
Random Forest Classification
Random forest, like its name implies, consists of
a large number of individual decision trees that
operate as an ensemble.
Each individual tree in the random forest spits
out a class prediction and the class with the
most votes become our model’s prediction.

9. Machine Learning Classification Cont..
Logistic Regression Classification
statistical model (also known as logit model) is often used
for classification and predictive
the outcome is a probability, the dependent variable is
bounded between 0 and 1

10. Steps of Programming and Running
Python language in Anaconda software and
Jupyter model
1. Reading the Dataset
2. Checking the Dataset if it has missing values
3. Importing classifier
4. Applying machine training on 80% of the dataset

11. Steps of Programming and Running
5. Making Prediction for the rest 20% of dataset
6. Measuring the accuracy
7. Drawing the confusion matrix
8. Comparing the results of all the used classifiers
9. Testing the machine using new unseen and unlabeled
dataset

12. Results
Decision Tree
Accuracy: 97.3186119873817
Confusion Matrix

13. Random Forest
Accuracy: 98.58044164037855
Confusion Matrix
Results

14. Results
Logistic Regression
Accuracy: 92.74447949526814
Confusion Matrix

15. Conclusions
➢Human can make a decision for something but with limited volume.
➢The machine intervention is required when the data are huge but
should be trained perfectly by the human.
➢ The supervised machine learning algorithms can be used for doing that
purpose through classifying the input data.
➢ In this project, three supervised machine learning classifiers (Decision
Tree, Random Forest, and Logistic Regression) applied on the same voice
dataset and we got different accuracy values.
➢The results sows the best one is the Random Forest algorithm with
accuracy 98.5, while the others are 97.7 for the Decision Tree algorithm,
and finally 92.7 for the Logistic Regression.
➢ The accuracy of classification differs according to the type and size of
dataset.

16. References
[1] Filipa M.B. Lã, N. Polo, D. Ardura, P. Coronado, M. B. Fiuza, A. M. Ramírez, “Acoustics”,
available from: https://unedvoicelab.com/acoustics/, accessed in 6 Dec 2022.
[2] K. BECKER, “Identifying the Gender of a Voice using Machine Learning”, available from:
https://www.primaryobjects.com, Jun 22, 2016, accessed in 6 Dec 2022.
[3] Scikit Learn Team, “Decision Tree”, available From: https://scikit-learn.org/, accessed
in 6 Dec, 2022.
[4] T. Yiu, “Understanding Random Forest”, available from:
https://towardsdatascience.com/, Jun 12, 2019, accessed in 6 Dec, 2022.
[5] S. Sperandei, “Understanding logistic regression analysis”, Biochemia Medica, 2014.
[6] H. Belyadi, A. Haghighat, “Machine Learning Guide for Oil and Gas Using Python”,
Elsevier, 2021.

17. THANK YOU