AClassiHonk: A System Framework to Annotate and Classify Vehicular Honk from Road Traffic

Abstract

Recent studies emphasize that vehicular honking contributes to over 50% of noise pollution in developing urban and suburban areas. Frequent honking negatively impacts health, road safety, and the environment. Recognizing and classifying different vehicle honks could offer valuable insights into environmental noise pollution. Existing research on outdoor sound classification and honk detection lacks the ability to classify honks based on vehicle types, limiting contextual information inference for locations, areas, or traffic. Therefore, it becomes imperative to design a system that can detect and classify honks of different types of vehicles from which we can infer some contextual information. In this paper, we have developed a novel framework AClassiHonk that performs raw vehicular honk sensing, data labeling and classifies the honk into three major groups, i.e., light-weight vehicles, medium-weight vehicles, and heavy-weight vehicles. We collected the raw audio samples of different vehicular honking based on spatio-temporal characteristics and converted them into spectrogram images. We have proposed a deep learning-based Multi-label Autoencoder model (MAE) for automated labeling of the unlabeled data samples, which provides 97.64% accuracy in contrast to existing deep learning-based data labeling methods. Further, we have used various pre-trained models, namely Inception V3, ResNet50, MobileNet, ShuffleNet, and proposed an Ensembled Transfer Learning model (EnTL) for vehicle honks classification and performed comparative analysis. Results reveal that EnTL exhibits the best performance compared to pre-trained models and achieves 96.72% accuracy in our dataset. In addition, we have identified a context of a location based on these classified honk signatures in a city.