Internet of Things. User-Centric IoT. First International Summit, IoT360 2014, Rome, Italy, October 27-28, 2014, Revised Selected Papers, Part I

Research Article

Multilingual Voice Control for Endoscopic Procedures

Download
299 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-19656-5_33,
        author={Sim\"{a}o Afonso and Isabel Laranjo and Joel Braga and Victor Alves and Jos\^{e} Neves},
        title={Multilingual Voice Control for Endoscopic Procedures},
        proceedings={Internet of Things. User-Centric IoT. First International Summit, IoT360 2014, Rome, Italy, October 27-28, 2014, Revised Selected Papers, Part I},
        proceedings_a={IOT360},
        year={2015},
        month={7},
        keywords={Automatic speech recognition Hidden Markov Models Pocketsphinx Sphinxtrain Endoscopic procedures},
        doi={10.1007/978-3-319-19656-5_33}
    }
    
  • Simão Afonso
    Isabel Laranjo
    Joel Braga
    Victor Alves
    José Neves
    Year: 2015
    Multilingual Voice Control for Endoscopic Procedures
    IOT360
    Springer
    DOI: 10.1007/978-3-319-19656-5_33
Simão Afonso1,*, Isabel Laranjo1,*, Joel Braga1,*, Victor Alves1,*, José Neves1,*
  • 1: University of Minho
*Contact email: simaopoafonso@gmail.com, isabel@di.uminho.pt, joeltelesbraga@gmail.com, valves@di.uminho.pt, jneves@di.uminho.pt

Abstract

In this paper it is present a solution to improve the current endoscopic exams’ workflow. These exams require complex procedures, such as using both hands to manipulate buttons and pressing a foot pedal at the same time, to perform simple tasks like capturing frames for posterior analysis. In addition to this downside, the act of capturing frames freezes the video. The developed software module was integrated with the device, a device for the acquisition, processing and storage of the endoscopic results It uses libraries developed by the PocketSphinx project to recognize a small amount of commands. The module was fine-tuned for the Portuguese language which presents some specific difficulties with speech recognition. It was obtained a Word Error Rate (WER) of 23.3 % for the English model and 29.1 % for the Portuguese one.