Exploring Classification of Vehicles Using Horn Sound Analysis: A Deep Learning-Based Approach

Vehicle classification in real-life traffic scenarios is crucial for road safety and traffic management. However, it is a complex task due to the diverse vehicle types, uncontrolled environments, and limitations of computational resources. Although visual-based models are highly accurate, they require high computational resources and have limitations. On the other hand, audio-based vehicle classification has unique strengths, making it ideal for continuous, real-time monitoring and minimizing computational load. This study proposes a new approach to vehicle classification using a 1D Convolutional Neural Network (CNN1D) by analyzing the sound of vehicle horns. It involved collecting 200 horn sounds from various vehicles and analyzing, scaling, and labeling them through Audacity. The Mel-Frequency Cepstral Coefficients (MFCCs) were used for preprocessing and feature extraction. The CNN1D model, trained on this data, accurately classifies vehicles like bikes, buses, cars, and CNG autos. The proposed model was compared to other models, including Long Short-Term Memory (LSTM), Support Vector Machines (SVM), and Artificial Neural Networks (ANN), for validation purposes. It demonstrates a well-balanced and precise vehicle classification, resulting in high precision, recall, F1-score, and accuracy of 95.12%. Moreover, the model performs better than other assessments in terms of recall, accuracy, and precision, further validating the usage of the model. This study's proposed model can be a potential solution for real-world vehicle classification and detection applications, providing a viable path for real-world applications. It may also help identify which vehicle is responsible for continuous sound pollution, leading to better traffic management and environmental protection.