Zaznava izgovorjene besede na sistemu Android

diplomsko delo

Elian Mugerli (Author), Octavian-Mihai Machidon (Mentor)

Abstract

V okviru diplomskega dela je implementirana aplikacija za prepoznavo izgovorjenih besed na sistemu Android. Prepoznava se izvaja s pomočjo lokalno shranjenega modela TensorFlow Lite na napravi. Model je naučen s pomočjo značilk MFCC, pridobljenih iz nabora zvočnih posnetkov. Faze delovanja si sledijo tako, da aplikacija najprej zajame zvok na vhodu naprave, ga nato obdela v značilke in na pridobljeni matriki opravi klasifikacijo. Tako dosežemo neprekinjeno prepoznavo besed. Postopek obdelave zvočnega signala v aplikaciji mora biti ekvivalenten postopku obdelave, ki je uporabljen v cevovodu za učenje. Model na testnih podatkih dosega natančnost 88.73%, medtem ko, storitev aplikacije pri uporabniškem testiranju dosega natančnost 82.23% na podatkih v realnem svetu.

Keywords

prepoznava besed;zaznava izgovorjenih besed;MFCC;Android;univerzitetni študij;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2023
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[E. Mugerli]
UDC:	004.5:004.934(043.2)
COBISS:	166292227
Views:	34
Downloads:	5
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Custom wake-word detection on Android
Secondary abstract:	As part of the thesis, an Android application for wake word recognition is implemented. Recognition is performed using a locally stored TensorFlow Lite model on the device. The model is trained using MFCCs obtained from a custom set of audio recordings. The application operates by initially capturing audio from the device's input, subsequently transforming it into features, and then conducting classification on the resulting matrix. This process enables us to achieve continuous word recognition. The processing in the application must be equivalent to the processing from the model training. The model achieves an accuracy of 88.73% on test data, while the application, based on user testing, is 82.23% accurate on real-world data.
Secondary keywords:	mobile application;word detection;MFCC;Android;computer science;diploma;Mobilne aplikacije;Avtomatsko prepoznavanje govora;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000468
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	40 str.
ID:	19921120