diplomsko delo
Miha Breznik (Author), Simon Dobrišek (Mentor)

Abstract

Razpoznavanje besednih ukazov se dandanes uporablja v številnih procesih in napravah. Številne naprave so zgrajene z mikroprocesorji, ki imajo majhno procesorsko moč, kar omejuje velikost nevronskega omrežja. V diplomski nalogi je predstavljen sistem razpoznavanja besednih ukazov z uporabo konvolucijskega nevronskega omrežja. Nevronsko omrežje je bilo naučeno z vzorci, ki bodo povzročili prekomerno prileganje. Lastnost prekomernega prileganja se je uporabila za zviševanje natančnosti razpoznavanja. Izvedena je bila primerjava razpoznavanja vzorcev znanega govorca in vzorcev neznanega govorca. Primerjava je pokazala pozitivno lastnost prekomernega prileganja pri razpoznavanju besednih ukazov.

Keywords

razpoznavanje govornih ukazov;strojno učenje;prekomerno prileganje;visokošolski strokovni študij;Aplikativna elektrotehnika;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FE - Faculty of Electrical Engineering
Publisher: [M. Breznik]
UDC: 004.934:004.85(043.2)
COBISS: 168706307 Link will open in a new window
Views: 33
Downloads: 2
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Embedded system for speech command recognition using a convolutional neural network model
Secondary abstract: Recognition of verbal commands is being used in numerous processes and devices. Numerous devices are built with microprocessors, which have small processing power, which limits the size of neural networks. The thesis presents a system for recognizing verbal commands using a convolutional neural network. The neural network was trained with patterns that will result in overfitting. The overfitting property was used to increase the recognition accuracy. A comparison of the recognition of patterns of a known speaker and patterns of an unknown speaker was performed. The comparison showed a positive overfitting property in recognition of verbal commands.
Secondary keywords: recognition of verbal commands;machine learning;overfitting;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000315
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za elektrotehniko
Pages: VII, 30 str.
ID: 19921216