Authors: E. Sophiya; S. Jothilakshmi
Addresses: Department of Computer Science and Engineering, Faculty of Engineering and Technology (FEAT), Annamalai University, Annamalainagar, India ' Department of Information Technology, Faculty of Engineering and Technology (FEAT), Annamalai University, Annamalainagar, India
Abstract: Humans are surrounded by a complex audio stream that carries meaningful information about our surroundings in day to day life. Hearing is one of the most important capabilities to identify and detect audio events in our surroundings. For example sounds such as ambulance siren, gunshot, baby cry, etc which require immediate action. Thus, automatic audio analysis is getting popular in recent years which have wide range of applications such as continuous monitoring for public safety, abnormal events, wildlife monitoring, healthcare, audio indexing and retrieval. The objective of the proposed system is to provide the event class and the event time boundaries between multiple events present in an audio. The proposed audio event detection is implemented with a deep learning model. The real time data are collected from major locations of an urban city. Audio events were recognised using signal processing techniques. The model is learned from Log Mel spectrogram features.
Keywords: audio processing; audio scene analysis; audio event detection; deep learning; deep convolutional neural network.
International Journal of Computer Aided Engineering and Technology, 2022 Vol.16 No.3, pp.328 - 343
Received: 05 Feb 2019
Accepted: 22 Jul 2019
Published online: 11 Apr 2022 *