Develop VoiceLab Automatic Speech Recognition System (ARM) based on Deep Neural Networks (DNN) and develop an innovative RAMD (Rapid Acoustic Model Development) data collection and training system to achieve super-human performance. (Q120773)

From EU Knowledge Graph
Revision as of 07:10, 21 October 2020 by DG Regio (talk | contribs) (‎Changed label, description and/or aliases in 1 language: translated_label)
Jump to navigation Jump to search
Project in Poland financed by DG Regio
Language Label Description Also known as
English
Develop VoiceLab Automatic Speech Recognition System (ARM) based on Deep Neural Networks (DNN) and develop an innovative RAMD (Rapid Acoustic Model Development) data collection and training system to achieve super-human performance.
Project in Poland financed by DG Regio

    Statements

    0 references
    4,704,949.0 zloty
    0 references
    1,129,187.76 Euro
    13 January 2020
    0 references
    6,772,330.0 zloty
    0 references
    1,625,359.2 Euro
    13 January 2020
    0 references
    69.47 percent
    0 references
    1 July 2016
    0 references
    30 June 2020
    0 references
    VOICELAB.AI SP. Z O.O.
    0 references
    0 references
    Projekt polega na zbudowaniu systemu ARM (Automatyczne Rozpoznawanie Mowy) VoiceLab w celu osiągnięcia najwyższej skuteczności rozpoznawania mowy na poziomie dorównującym człowiekowi w analogicznych warunkach akustycznych (super human performance) w oparciu o głębokie sieci neuronowe DNN (Deep Neural Network). Nowy system ARM VoiceLab będzie działał w trudnych warunkach akustycznych w oparciu o bardzo dużą ilości danych treningowych, dla różnych typów języków, przy uzyskaniu poziomu wyników konkurencyjnych do najlepszych tego typu rozwiązań na świecie. Równorzędnym celem projektu jest również zbudowanie systemu do szybkiego zbierania danych wykorzystywanych do treningu modeli akustycznych RAMD (Rapid Acoustic Model Development). System ten będzie umożliwiał iteracyjnie prowadzenie badań i kontynuowanie dalszych prac rozwojowych nad rozpoznawaniem mowy. (Polish)
    0 references
    The project consists of building a VoiceLab (Automatic Speech Recognition) system in order to achieve the highest efficiency of speech recognition at a level equal to humans under super human performance based on deep neural networks DNN (Deep Neural Network). The new ARM VoiceLab system will operate in difficult acoustic conditions based on a very large amount of training data, for different types of languages, with a level of competitive results to the best of this type in the world. The equivalent goal of the project is also to build a system for quick data collection used for training RAMD acoustic models (Rapid Acoustic Model Development). This system will enable iterative research and further development of speech recognition. (English)
    21 October 2020
    0 references

    Identifiers

    RPPM.01.01.01-22-0026/16
    0 references