Machine Learning-Based Identification of Stock Car Models via Car Sound Analysis
| dc.contributor.author | Isaiah Ho Chi Ann | |
| dc.date.accessioned | 2026-04-29T03:31:57Z | |
| dc.date.issued | 2025 | |
| dc.description | The identification of vehicle models traditionally relies on visual recognition methods, which can be hindered by poor lighting, physical obstructions, or adverse weather conditions. This project introduces an innovative approach to vehicle model identification by leveraging machine learning techniques and car sound analysis. Each stock car model emits a unique acoustic signature influenced by its engine configuration and exhaust design. By analysing these sound patterns, the proposed system aims to classify stock car models with high accuracy. Through the utilisation of modern machine learning models, they have the power and intelligence to easily and most importantly quickly identify stock car models with verifiable accuracy. A comprehensive dataset of 233 stock car models was created using high-quality audio recordings from the game Forza Horizon 5. A robust pipeline was implemented for audio preprocessing and feature extraction, utilizing Mel-Frequency Cepstral Coefficients (MFCCs), spectral centroid, and other tonal features to represent the acoustic signatures numerically. Three machine learning models were developed and evaluated: Support Vector Machine (SVM), Random Forest (RF), and a Convolutional Neural Network (CNN). Evaluation using 5-fold cross-validation demonstrated high performance, with the SVM (95.72%) achieving the highest mean accuracy, followed by CNN (95.28%) and RF (93.02%). While the system works well on the controlled dataset, its accuracy drops when tested on real-world audio samples, highlighting the need for better generalization. This project successfully demonstrates the viability of acoustic-based classification in controlled settings and contributes to the advancement of non-visual vehicle identification techniques. The findings highlight key limitations and pave the way for future research. | |
| dc.identifier.uri | https://scholarhub.unimas.my/handle/123456789/560 | |
| dc.language.iso | English | |
| dc.publisher | Universiti Malaysia Sarawak | |
| dc.relation.ispartofseries | Faculty of Computer Science and Information Technology | |
| dc.title | Machine Learning-Based Identification of Stock Car Models via Car Sound Analysis | |
| dc.type | Final Year Project |