Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

An Open-Set Recognition and Few-Shot Learning Dataset for Audio Event Classification in Domestic Environments

  • Naranjo-Alcaraz, Javier
  • Perez-Castanos, Sergi
  • Zuccarello, Pedro
  • Torres, Ana M.
  • Ferri, Francesc J.
  • Cobos, Maximo
  • López Monfort, José Javier|||0000-0001-6884-5577
[EN] The problem of training with a small set of positive samples is known as few-shot learning (FSL). It is widely known that traditional deep learning algorithms usually show very good performance when trained with large datasets. However, in many applications, it is not possible to obtain such a high num- ber of samples. This paper deals with the application of FSL to the detection of specific and intentional acoustic events given by different types of sound alarms, such as door bells or fire alarms, using a limited number of samples. These sounds typically occur in domestic environments where many events corre- sponding to a wide variety of sound classes take place. Therefore, the detection of such alarms in a practi- cal scenario can be considered an open-set recognition (OSR) problem. To address the lack of a dedicated public dataset for audio FSL, researchers usually make modifications on other available datasets. This pa- per is aimed at providing the audio recognition community with a carefully annotated dataset1 for FSL in an OSR context comprised of 1360 clips from 34 classes divided into pattern sounds and unwanted sounds. To facilitate and promote research on this area, results with state-of-the-art baseline systems based on transfer learning are also presented., This work was supported by the EU Horizon 2020 programme [grant No 779158] . Grants DIN2018-009982, PTQ-17-09106, RTI2018-097045-B-C21/C22 funded by MCIN/AEI/10.13039/50110 0 011033, the latter also by "ERDF A way of making Europe ". Grants TED2021-131003B-C21/C22 funded by MCIN/AEI/10.13039/50110 0 011033 and by the "EU Union NextGenerationEU/PRTR ". Grants AICO/2020/154 and AEST/2020/012, funded by GVA. The authors acknowledge also the Artemisa computer resources funded by the EU ERDF and Comunitat Valenciana, and the technical support of IFIC (CSIC-UV) . Authors J. Naranjo, S. Perez and P. Zuccarello were working at Visualfy when this work was done, but they are now with the Instituto Tecnologico de Informatica (ITI) , Tyris AI and ITI, respectively.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Enhancing Road Safety with AI-Powered System for Effective Detection and Localization of Emergency Vehicles by Sound

  • Banchero-Martínez, Lucas
  • Vacalebri-Lloret, Francisco
  • Mossi García, José Manuel|||0000-0001-9083-3476
  • López Monfort, José Javier|||0000-0001-6884-5577
[EN] This work presents the design and implementation of an emergency sound detection and localization system, specifically for sirens and horns, aimed at enhancing road safety in automotive environments. The system integrates specialized hardware and advanced artificial intelligence algorithms to function effectively in complex acoustic conditions, such as urban traffic and environmental noise. It introduces an aerody-namic structure designed to mitigate wind noise and vibrations in microphones, en-suring high-quality audio capture. In terms of analysis through artificial intelligence, the system utilizes transformer-based architecture and convolutional neural networks (such as residual networks and U-NET) to detect, localize, clean, and analyze nearby sounds. Additionally, it operates in real-time through sliding windows, providing the driver with accurate visual information about the direction, proximity, and trajectory of the emergency sound. Experimental results demonstrate high accuracy in both con-trolled and real-world conditions, with a detection accuracy of 98.86% for simulated data and 97.5% for real-world measurements, and localization with an average error of 5.12° in simulations and 10.30° in real-world measurements. These results highlight the effectiveness of the proposed approach for integration into driver assistance systems and its potential to improve road safety., This project was supported by the project AEOLIAN:OMAP (TED2021-131003B-C22) in PTRS by the Spanish Government.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Advancements in Vertical Jump Measurement Using Sound and Artificial Intelligence: Precision and Portability, Avances en la medición de saltos verticales mediante sonido e Inteligencia Artificial: precisión y portabilidad

  • Banchero-Martínez, Lucas
  • López Monfort, José Javier|||0000-0001-6884-5577
[EN] Accurate measurement of vertical jump height is crucial for assessing muscular performance and body coordination in athletes and active individuals. Traditional methods such as force platforms and motion capture systems have been the standard but pose significant challenges in terms of portability and technical complexity. In response to these limitations, mobile devices have garnered attention as an accessible and effective alternative. This article explores recent advances in jump measurement using audio technology and Artificial Intelligence (AI) techniques. An algorithm based on deep neural networks has been developed to accurately identify key take-off and landing events from sound recordings. This approach overcomes inherent challenges of environmental noise in sports settings, demonstrating high precision and reliability compared to conventional methods. Furthermore, a new approach is being investigated to optimize the experimental setup for increased practicality and accessibility, without the need to add additional elements on athletes. These advancements promise not only to enhance measurement accuracy but also to facilitate implementation in sports and physical rehabilitation settings. This study underscores the transformative potential of AI in evaluating physical performance, providing innovative and effective tools for sports and health professionals., [ES] La medición precisa de la altura en saltos verticales es crucial para evaluar el rendimiento muscular y la coordinación corporal en deportistas y personas activas. Los métodos tradicionales, como las plataformas de fuerza y los sistemas de captura de movimiento, han sido el estándar, pero presentan desafíos significativos en términos de portabilidad y complejidad técnica. En respuesta a estas limitaciones, el uso de dispositivos móviles ha ganado atención como una alternativa accesible y efectiva. Este artículo explora avances recientes en la medición de saltos utilizando tecnología de audio y técnicas de Inteligencia Artificial (IA). Se ha desarrollado un algoritmo basado en redes neuronales profundas que permite identificar con precisión los eventos clave de despegue y aterrizaje a partir de grabaciones de sonido. Este enfoque supera los desafíos inherentes al ruido ambiental en entornos deportivos, demostrando una alta precisión y fiabilidad comparada con métodos convencionales. Además, se está investigando una nueva aproximación que optimiza el set-up experimental para hacerlo más práctico y accesible, sin la necesidad de añadir elementos adicionales en los atletas. Estos avances prometen no solo mejorar la precisión de las mediciones, sino también facilitar su implementación en entornos deportivos y de rehabilitación física. Este estudio subraya el potencial transformador de la IA en la evaluación del rendimiento físico, proporcionando herramientas innovadoras y eficaces para profesionales del deporte y la salud., Este proyecto ha sido subvencionado con los proyectos SONEM3D (CIAICO/2021/057) y AEOLIAN:OMAP (TED2021-131003B-C22).