Research on Urban Environmental Sound Classification Model Based on Multi-Convolutional Neural Networks

With the rapid acceleration of urbanization, the impact of urban environmental sounds is garnering increasing attention. This study aims to develop a classification system capable of swiftly identifying various types of urban environmental sounds, thereby facilitating audio classification and noise pollution management. The research employs a Multi-Convolutional Neural Network (MULT-CNN) algorithm, which enhances the accuracy of audio classification by averaging the results of multiple CNN models. The study utilizes the widely recognized Urbansound8K dataset, which encompasses ten distinct types of urban sounds. Through model training and testing, the system's performance was evaluated using metrics such as accuracy, precision, recall, F1-Score, and processing latency. Experimental results demonstrate that the classification model performs well across most categories, particularly in Category 5 (Engine Idling) and Category 2 (Children Playing), where it achieves high classification accuracy and low processing latency. Additionally, audio classification heatmaps and confusion matrix heatmaps were developed to visually represent the model's effectiveness in audio classification. This research provides an effective technical approach for urban sound classification and management.