bachelor's thesis

Abstract

A time series is a sequence of points ordered in time. Time series analysis can often reveal useful patterns for describing certain behavior or for predicting future events. In this thesis, we experimentally express the relationship between the symptoms severity scores of the patients and their gait signals defined as time series. We used different deep neural networks for time series classification and investigated the ability of deep neural networks to automatically extract discriminatory features from raw sensory data. We show how transferred features from the bottom, middle, or top layer of the neural network for human activity recognition affect the models' performance for detection of the symptoms. We empirically assess the accuracy of deep neural networks in a practical scenario where we try to automatically predict the patients' symptoms based on their gait signals.

Keywords

wearable sensors;Parkinson's disease;deep learning;computer and information science;diploma;

Data

Language: English
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [A. Petrushevski]
UDC: 004.8:616.858(043.2)
COBISS: 1538421187 Link will open in a new window
Views: 735
Downloads: 294
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: Slovenian
Secondary title: Detekcija simptomov Parkinsonove bolezni na podlagi nosljivih senzorjev
Secondary abstract: Časovna vrsta je zaporedje časovno razporejenih točk. Analiza časovnih vrst pogosto razkrije uporabne vzorce za opis določenih vedenj ali za napovedovanje prihodnjih dogodkov. Namen diplomskega dela je eksperimentalno določiti resnost simptomov bolnikov z meritvami, pridobljenimi iz pospeškometra in žiroskopa, ki so definirane kot časovne vrste. Za klasifikacijo časovnih vrst smo uporabili globoke nevronske mreže. Raziskali smo sposobnost globokih nevronskih mrež, da samodejno pridobivajo diskriminatorne lastnosti iz surovih senzoričnih podatkov. Pokažemo, kako značilke iz začetnih, sredinskih in končnih nivojev mreže za prepoznavanje človeške dejavnosti vplivajo na uspešnost modelov za odkrivanje simptomov parkinsonove bolezni. Empirično preverimo natančnost globokih nevronskih mrež v praktičnem scenariju, kjer poskušamo oceniti bolnikove simptome na podlagi signalov nosljivih senzorjev med hojo.
Secondary keywords: nosljivi senzorji;konvolucijske nevronske mreže;Parkinsonova bolezen;globoko učenje;računalništvo in informatika;univerzitetni študij;diplomske naloge;
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: 61 str.
ID: 11255027