Real-time active-learning method for audio-based anomalous event identification and rare events classification for audio events detection

Introduction: Audio event detection, the application of scientific methods to analyze audio recordings, can be helpful in examining and analyzing audio recordings to preserve, analyze, and interpret sound evidence. Furthermore, it can be helpful in safety and compliance, security, surveillance, maintenance, and predictive analysis. Audio event detection aims to recover meaningful information from audio recordings, such as determining the authenticity of the recording, identifying the speakers, and reconstructing conversations. However, filtering out noise for better accuracy in audio event detection is a major challenge. A greater sense of public security can be achieved by developing automated event detection systems that are both cost-effective and real-time. Methods: In response to these challenges, this study presented a method for identifying anomalous events based on noisy audio evidence and a real-time scenario to help the audio event detection investigator during the investigation. This study created a large audio dataset containing both noisy and original audio. The dataset includes diverse environmental background settings (e.g., office, restaurant, and park) and some abnormal events (e.g., explosion, car crash, and human attack). This study used an ensemble learning model to conduct experiments in an active learning environment. Nine methods are employed to create the feature vector. Results: The experiments show that the proposed ensemble learning model using the active learning settings obtained an accuracy score of 99.26%, while the deep learning model obtained an accuracy of 95.35%. The proposed model was tested using noisy audio evidence and a real-time scenario. Discussion: The experiment results show that the proposed approach can efficiently detect abnormal events from noisy audio evidence and a real-time scenario in real-time.