Master’s Thesis at the College of Information Technology Explores Signal Classification Using DCT and Deep Learning

Duhaa Fadill Abbas
The College of Information Technology at the University of Babylon hosted a master’s thesis defense in the Software Department titled: “Classification of Signal Types Based on DCT and Deep Learning ”, presented by graduate student Haitham Mohammed Abbas Shakir, and supervised by Prof. Dr. Tawfiq Abdulkhaleq Abbas. The discussion took place on Wednesday, July 9th, 2025, in the college’s Conference Hall, in the presence of faculty members, researchers, and postgraduate students.

The thesis highlighted the vital role of sound analysis in real-world applications, particularly in industrial and veterinary fields. It emphasized how experienced operators can often detect mechanical faults in rotating machinery by auditory cues. However, in noisy industrial environments, fault detection becomes significantly more challenging. Likewise, poultry and livestock farmers can recognize disease symptoms and behavioral changes in animals by analyzing their vocal patterns.

The research proposed an intelligent system for classifying and analyzing audio signals using deep learning techniques, specifically a Convolutional Neural Network (CNN). The system first converts audio signals into grayscale images during the pre-processing phase, where background noise is suppressed using the Discrete Cosine Transform (DCT). These transformed images are then fed into the CNN to perform multi-class classification.

The experimental results demonstrated outstanding performance, achieving 100% accuracy in detecting motor faults, 97.1% in identifying animal sounds, and 97.6% in classifying poultry vocalizations. Moreover, the model maintained high robustness across diverse noise conditions in audio recordings.

This study constitutes a significant contribution to the field of audio signal analysis, paving the way for innovative technologies that can be applied across scientific, industrial, and medical domains. It showcases the effectiveness of integrating signal processing techniques with artificial intelligence to enhance sound recognition and interpretation in complex environments.