Επιτάχυνση ενός αλγόριθμου διαχωρισμού ομιλίας/ μουσικής με χρήση FPGA
FPGA-based acceleration of a speech / music discrimination algorithm
View/
Subject
Επεξεργασία σημάτων ; Field programmable gate arrays -- Design and construction ; Signal processing ; Signal processing -- Digital techniquesAbstract
This thesis objective is the implementation of a speech-music discrimination algorithm. This refers to the identification and classification of audio stream segments - or else frames - as speech or music. The implemented method is based on extracting from the audio recording, a feature called spectral entropy. As soon as spectral entropy has been extracted from an audio frame, mean value and standard deviation are calculated. Frames are classified depending on standard deviation's value, and specifically whether it is bigger than a predefined threshold. The method was implemented on Xilinx ML403 development board, which includes an FPGA of the Virtex 4 family. The developed system is based on the PowerPC processor that resides within the FPGA. Parts of the algorithm, and specifically audio frames sampling and frequency spectrum calculation are implemented directly in hardware for performance reasons.
Postgraduate Studies Programme
Προηγμένα Συστήματα ΠληροφορικήςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
68Language
GreekCollections
Except where otherwise noted, this item's license is described as
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές