Robustnost govornega vmesnika Google Home v hrupnem okolju

diplomsko delo

Matevž Praček (Author), Marko Meža (Mentor)

Abstract

Diplomsko delo poskuša odgovoriti na vprašanja, v kakšnih pogojih in v kakšnem razmerju hrupa nas Google Home Mini sliši in se na to odzove. Predvsem, kako se Google Home Mini govorni vmesnik obnese v vsakdanjem življenju. Ali bo razumel samo eno osebo, če se več ljudi pogovarja v ozadju, ali bo ubogal le naše ukaze, in do kakšne jakosti šuma v ozadju se odzove. V diplomskem delu je predstavljen eksperiment, pri katerem se spreminja glasnost posnetkov govora in posnetkov šuma v zaprtem okolju. V drugem delu eksperimenta je prikazan izračun razmerja signal-šum in ugotovitev, da je meja razumevanja v hrupnem okolju dokaj nizka in da potrebuje tišje okolje. Na primer, če je šum jakosti 79 dB in izgovorjena fraza jakosti 77 dB je malo verjetno, da se Google Home Mini odzove na ukaz. Četudi ga v dveh od štirih primerov zazna, se nanj ne odzove.

Keywords

govorni vmesniki;Google Home Mini;hrupno okolje;robustnost;visokošolski strokovni študij;Multimedijske komunikacije;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2021
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FE - Faculty of Electrical Engineering
Publisher:	[M. Praček]
UDC:	004.512.4(043.2)
COBISS:	91403267
Views:	121
Downloads:	12
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Robustness of speach-based user interface Google Home in a noisy environment
Secondary abstract:	The thesis tries to answer the questions in what conditions and in what noise ratio Google Home Mini hears us and responds to it. Especially how the Google Home Mini voice interface works in everyday life. Will he understand only one person if more people are talking in the background, or will he just obey our commands, and to what intensity of noise in the background does he respond. The thesis presents an experiment in which the volume of speech recordings and noise recordings in a closed environment are changed. The second part of the experiment shows the calculation of the signal-to-noise ratio and the finding that the limit of understanding in a noisy environment is quite low and that it needs a quieter environment. For example, if the noise is 79 dB and the spoken phrase is 77 dB, it is unlikely that the Google Home Mini will respond to the command. Even if he detects it in two of the four cases, he does not respond to it.
Secondary keywords:	Google;Home;Mini;speach;based;interface;noisy;environment;robustness;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000314
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za elektrotehniko
Pages:	61 str.
ID:	14119250