Advancements in Environmental Sound Classification: Evaluating Machine Learning and Deep Learning Approaches on the UrbanSound8k
Priyanshu Malaviya, Yogesh Kumar, Nandini Modi · 2023
In the rapidly evolving domain of audio classification, the quest for optimal model performance remains paramount. This research embarked on a comprehensive journey, juxtaposing traditional machine learning algorithms with state-of-the-art deep learning architectures, all benchmarked on the UrbanSound8k dataset. The study meticulously evaluated models based on pivotal metrics: Accuracy, Precision, Recall, and F1-score. Traditional models, including RandomForest and KNeighbors Classifier, showcased promising results, with the latter achieving a remarkable accuracy surge post hyperparameter tuning. However, the deep learning models, particularly the Artificial Neural Network (ANN), emerged as the zenith, registering an astounding accuracy of 97.59% after optimization. This paper not only underscores the prowess of deep learning in audio classification but also emphasizes the significance of hyperparameter refinement. The findings presented herein offer invaluable insights, setting the stage for future endeavors in the realm of audio data analysis.