CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Start-and-End Point Detection at the Input of Speech Recognition Application

Estudiants que han llegit aquest projecte:

Director/a: TARRÉS RUIZ, FRANCESC

Departament: TSC

Títol: Start-and-End Point Detection at the Input of Speech Recognition Application

Data inici oferta: 29-05-2013     Data finalització oferta: 29-05-2013


Estudis d'assignació del projecte:
    Tipus: Individual
     
    Lloc de realització: ERASMUS
     
            Supervisor/a extern: Ptr Pollack
            Institució/Empresa: CTU - Prague
            Titulació del Director/a: Telecom Engineer
     
    Paraules clau:
    Voice Activity Detection, Speech Processing
     
    Descripció del contingut i pla d'activitats:
    The objective of the project is the creation of an algorithm for Start-and-End point
    detection of a pre-recorded signal.
    The initial reason for developing this algorithm is so it can be used at the input of a
    voice recognition application. Overall, the result of this work is an algorithm that can
    detect the beginning and end of a previously recorded signal based on a detection
    algorithm of the voice activity previously developed by the Czech Technical University,
    Faculty of Electrical Engineering.
    Two main issues are studied in this project: Detecting the Voice Activity (VAD
    algorithm) and determining the start and end point of the signal (Start-and-End point
    detection).
    To demonstrate the mode of operation of the algorithm, I have created an application
    in MATLAB to show graphically the process for a previously recorded signal and then
    the start and end points after applying the algorithms.
    Finally, to provide better graphic performance and provide added value to the project, I
    have added a digit recognition algorithm based on a DTW (Dynamic Time Warping).

     
    Overview (resum en anglès):

    The objective of the project is the creation of an algorithm for Start-and-End point
    detection of a pre-recorded signal.
    The initial reason for developing this algorithm is so it can be used at the input of a
    voice recognition application. Overall, the result of this work is an algorithm that can
    detect the beginning and end of a previously recorded signal based on a detection
    algorithm of the voice activity previously developed by the Czech Technical University,
    Faculty of Electrical Engineering.
    Two main issues are studied in this project: Detecting the Voice Activity (VAD
    algorithm) and determining the start and end point of the signal (Start-and-End point
    detection).
    To demonstrate the mode of operation of the algorithm, I have created an application
    in MATLAB to show graphically the process for a previously recorded signal and then
    the start and end points after applying the algorithms.
    Finally, to provide better graphic performance and provide added value to the project, I
    have added a digit recognition algorithm based on a DTW (Dynamic Time Warping).

    © CBLTIC Campus del Baix Llobregat - UPC