College of Computer Science and Mathematics

Defense of a master’s thesis in the College of Computer Science and Mathematics – Department of Computer Science Discussed College of Computer Science and Mathematics at the University of Mosul Master’s thesis entitled: Automatic classification of smokers from the speech signal Thursday 14/3/2024

The thesis submitted by the student Ibrahim Khader Zaal Majnan in the Department of Halsoub Sciences dealt with the scientific material

The study aims to build a machine learning system to classify people who smoke from non-smokers through speech signal and using MFCC coefficients.

The study dealt with the classification of people who smoke from non-smokers through the speech signal (human voice)

A system is designed to classify non-smokers and their gender based on the speech signal using Python’s machine learning algorithms. This study focuses on the automatic classification of smokers using speech signals.

After preprocessing the dataset and analyzing the signals to extract the features, two types of features are extracted: Cepstral frequency inclination coefficients (MFCC) and power features (F0, jitter, flash, and HNR) extracted from the signals. The backpropagation neural network, support vector device (SVM), and decision tree cascade model (DT) begin training after data set preprocessing. Data classification is one of the main uses of machine learning algorithms.

The performance of the models was evaluated using the confusion matrix algorithm, which was achieved by the backpropagation network model (accuracy = 0.83 for males and accuracy = 0.82 for females) achieved the SVM model (accuracy = 0.96 for male accuracy = 0.93 for females), and the decision tree model (DT) was achieved (accuracy = 0.75 for males and accuracy = 0.83 for females). The SVM algorithm has achieved the highest rating rate, which is to correctly classify 4 people out of 5 people.

The discussion committee was chaired by Prof. Dr. Fawzia Mahmoud Remo and the membership of Assistant Prof. Dr. Samaa Tali Aziz

Dr. Omar Moayad Abdullah

Under the supervision and membership of the Assistant Professor

Yusra Faisal Mohammed