Javascript must be enabled for the correct page display

Swimmer Classificationusing an Artificial Lateral Lineand Time Series Approaches

Lambert, Daan (2022) Swimmer Classificationusing an Artificial Lateral Lineand Time Series Approaches. Master's Thesis / Essay, Artificial Intelligence.

[img]
Preview
Text
Master_s_Thesis_Daan_Lambert_s2974576.pdf

Download (3MB) | Preview
[img] Text
toestemming.pdf
Restricted to Registered users only

Download (120kB)

Abstract

The lateral line system is a sensory organ most commonly found in aquatic vertebra. These systemstransduce movement of the water into a stimulus. Artificial analogues of such systems can be con-structed to replicate this conversion from movement to signal. In this thesis the signal produced byan artificial lateral line (ALL) is used for classification purposes. Recordings were made of 5 swim-mers passing by an ALL. We showed success in identifying the swimmers through the utilisationof several classification techniques. We tested two transformation techniques as data representationsalong side the raw data. The two data transformations which were tested are the spectrogram, and thediscrete wavelet transform of the data. Both of these data representations incorporate the frequencydomain as well as the time domain. For the classification we used three different machine learningapproaches. These are a nearest neighbours approach, a time series forests approach, and a deeplearning approach. All three of these approaches represent a different niche in the field of time seriesclassification. We show that the raw data representation outperforms the spectrogram in all situations.The wavelet transform outperforms the raw data depending on the classifier used. The classifiers pro-duce similar results on the data set, when comparing the best performing data representations. Eachof the classifiers can be useful depending on circumstances and aim of the system. Future researchs

Item Type: Thesis (Master's Thesis / Essay)
Supervisor name: Netten, S.M. van and Wolf, B.J.
Degree programme: Artificial Intelligence
Thesis type: Master's Thesis / Essay
Language: English
Date Deposited: 28 Jan 2022 10:33
Last Modified: 28 Jan 2022 10:33
URI: https://fse.studenttheses.ub.rug.nl/id/eprint/26507

Actions (login required)

View Item View Item