Please use this identifier to cite or link to this item:
Title: A Sound Events Detection and Localization System based on YAMNet Model and BLE Beacons
Authors: Mesa-Cantillo, Carlos M.
Alonso-González, Itziar G. 
Quintana-Suárez, Miguel A. 
Ley-Bosch, Carlos 
Ramírez-Casañas, Carlos 
Sánchez-Medina, Javier J. 
Sánchez-Rodríguez, David 
UNESCO Clasification: 332505 Radiocomunicaciones
Keywords: sound event classification
machine learning
Issue Date: 2023
Journal: International Conference on Wireless and Mobile Communications (19.Barcelona.2023)
Abstract: Automatic recognition of sound events by computers is a requirement for several emerging applications, such as surveillance, automatic listening, and noise source identification. Acoustic Event Detection (AED) has the objective of knowing the identity of sounds and their temporal position in signals captured by one or several microphones. In this work, a pretrained Yet Another Mobile Network (YAMNet) model is used to perform real-time audio classification. This model is an audio event classifier that takes the audio waveform as input and makes independent predictions for each of the 521 audio events in the AudioSet ontology. The model uses the MobileNet v1 architecture and was trained using the AudioSet corpus. By utilizing a Raspberry Pi 3, a commercial microphone and a set of beacons, this system detects when an event occurs. Thus, the system can detect where and what event has been detected and send it to a database. After that, an update is made in the database and a notification can be sent to the users of a specific application. This information may be useful for people with disabilities to be warned of danger in nearby areas.
ISSN: 2308-4219
Appears in Collections:Artículos
Adobe PDF (206,28 kB)
Show full item record

Google ScholarTM



Export metadata

Items in accedaCRIS are protected by copyright, with all rights reserved, unless otherwise indicated.