Tiny Neural Network Pipeline for Vocal Commands Recognition @Edge

Ivana Guarneri, Alessandro Lauria, Giovanni Maria Farinella, Corrado Santoro

2022

Abstract

Vocal analysis and speech recognition have been a challenge for the research community for a long time. The widespread of deep learning approaches, the availability of big datasets and the increasing computational capabilities of processors, have contributed to achieve disruptive results in this field. Most of the high performing existing speech recognition systems are available as cloud services. Other systems are hybrid, with some parts on the cloud and some modules running on the microcontroller. One of the challenges is to realize high performing speech recognition systems running on the edge, where the edge is an integrated platform, composed by a processing unit, a bank of memory and a power unit. In this paper is proposed an end-to-end deep learning approach to recognize a set of vocal commands able to work on an edge IoT node. Tests have been performed on a tiny platform and the study with different settings is reported.

Paper Citation

in Harvard Style

Guarneri I., Lauria A., Farinella G. and Santoro C. (2022). Tiny Neural Network Pipeline for Vocal Commands Recognition @Edge. In Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 2: HUCAPP; ISBN 978-989-758-555-5, SciTePress, pages 249-254. DOI: 10.5220/0010908800003124

in Bibtex Style

@conference{hucapp22,
author={Ivana Guarneri and Alessandro Lauria and Giovanni Maria Farinella and Corrado Santoro},
title={Tiny Neural Network Pipeline for Vocal Commands Recognition @Edge},
booktitle={Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 2: HUCAPP},
year={2022},
pages={249-254},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010908800003124},
isbn={978-989-758-555-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 2: HUCAPP
TI - Tiny Neural Network Pipeline for Vocal Commands Recognition @Edge
SN - 978-989-758-555-5
AU - Guarneri I.
AU - Lauria A.
AU - Farinella G.
AU - Santoro C.
PY - 2022
SP - 249
EP - 254
DO - 10.5220/0010908800003124
PB - SciTePress