Applications of Signal Processing to Microphone Node Calibration and Medical Signal Classification
(2017)- Abstract
- Localization is an important enabling technology for many applications, such as wireless sensor networks, emergency rescue services, civil defense and transportation. Suppose that a room is equipped with several microphones (or sensors), and one person is making a sound while moving around in the room. Can one find microphone and sound source positions as well as reconstruct a room geometry? The answer is “Yes”. In the first part of the thesis, a system for microphone self-localization based on ambient sound without any assumptions on the 3D locations of the microphones and sound sources has been developed. The results show that the system works well when there is one dominant sound source. For such a setup the resulting accuracy is in the... (More)
- Localization is an important enabling technology for many applications, such as wireless sensor networks, emergency rescue services, civil defense and transportation. Suppose that a room is equipped with several microphones (or sensors), and one person is making a sound while moving around in the room. Can one find microphone and sound source positions as well as reconstruct a room geometry? The answer is “Yes”. In the first part of the thesis, a system for microphone self-localization based on ambient sound without any assumptions on the 3D locations of the microphones and sound sources has been developed. The results show that the system works well when there is one dominant sound source. For such a setup the resulting accuracy is in the range of few millimeters. In addition, we also considered the problem when two microphones are fixed on a rigid rack. For such a constellation, particularly the minimal problems (i.e. minimal number of microphones and sources needed to solve the problem) have been analyzed and we have identified the number of solutions.
Preterm birth is among the top causes of death in infants worldwide. Nowadays, increasing numbers of extremely preterm infants are surviving, however, they are at greater risk for short and long-term complications. The problem examined in the thesis is to analyze electroencephalogram (EEG) activity burst patterns to predict outcome in preterm infants (fetus born between 23-30 weeks gestational age) by using feature extraction and machine learning techniques. For this purpose, pilot datasets from 15 extremely preterm born infants have been studied. The results showed 80% of precision on the test set. However, because of the limited dataset, it is not possible to draw a solid conclusion about the exact predictability of our method.
Nowadays surveillance of fetal condition during labor is made using cardiotocography (CTG). While there are impressive achievements in modern obstetric care, there is still room for improving automatic bed-side CTG-interpretation for finding early signs of fetal distress. Considerable expertise is required to interpret whether the fetal response to the uterine contractions is adequate, or fetal response shows sign of fetal exhaustion or asphyxia. In this work, we aim to automatically detect suspicious patterns in short intervals by finding the correlation between fetal heart rate patterns and uterine contraction patterns. Our proposed method has shown some predictive power, however, further research is needed to improve the results for this important problem.
(Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/6b2f0a53-73ac-44d8-bb2a-a295f8d73d80
- author
- Simayijiang, Zhayida LU
- supervisor
-
- Karl Åström LU
- Carl Olsson LU
- opponent
-
- Professor Kjellström, Hedvig, , KTH, Stockholm
- organization
- publishing date
- 2017-01-11
- type
- Thesis
- publication status
- in press
- subject
- keywords
- ToA, TDoA, Sensor Network Localization, EEG, CTG, machine learning.
- pages
- 189 pages
- publisher
- Lund University, Faculty of Science, Centre for Mathematical Sciences, Mathematics
- defense location
- lecture hall MA:06, Matematikhusets Annex, Sölvegatan 18, Lund University, Faculty of Engineering, Lund
- defense date
- 2017-02-10 10:15:00
- ISBN
- 978-91-7753-078-7
- 978-91-7753-079-4
- language
- English
- LU publication?
- yes
- id
- 6b2f0a53-73ac-44d8-bb2a-a295f8d73d80
- date added to LUP
- 2017-01-13 15:45:44
- date last changed
- 2022-04-08 15:27:37
@phdthesis{6b2f0a53-73ac-44d8-bb2a-a295f8d73d80,
abstract = {{Localization is an important enabling technology for many applications, such as wireless sensor networks, emergency rescue services, civil defense and transportation. Suppose that a room is equipped with several microphones (or sensors), and one person is making a sound while moving around in the room. Can one find microphone and sound source positions as well as reconstruct a room geometry? The answer is “Yes”. In the first part of the thesis, a system for microphone self-localization based on ambient sound without any assumptions on the 3D locations of the microphones and sound sources has been developed. The results show that the system works well when there is one dominant sound source. For such a setup the resulting accuracy is in the range of few millimeters. In addition, we also considered the problem when two microphones are fixed on a rigid rack. For such a constellation, particularly the minimal problems (i.e. minimal number of microphones and sources needed to solve the problem) have been analyzed and we have identified the number of solutions.<br/>Preterm birth is among the top causes of death in infants worldwide. Nowadays, increasing numbers of extremely preterm infants are surviving, however, they are at greater risk for short and long-term complications. The problem examined in the thesis is to analyze electroencephalogram (EEG) activity burst patterns to predict outcome in preterm infants (fetus born between 23-30 weeks gestational age) by using feature extraction and machine learning techniques. For this purpose, pilot datasets from 15 extremely preterm born infants have been studied. The results showed 80% of precision on the test set. However, because of the limited dataset, it is not possible to draw a solid conclusion about the exact predictability of our method.<br/>Nowadays surveillance of fetal condition during labor is made using cardiotocography (CTG). While there are impressive achievements in modern obstetric care, there is still room for improving automatic bed-side CTG-interpretation for finding early signs of fetal distress. Considerable expertise is required to interpret whether the fetal response to the uterine contractions is adequate, or fetal response shows sign of fetal exhaustion or asphyxia. In this work, we aim to automatically detect suspicious patterns in short intervals by finding the correlation between fetal heart rate patterns and uterine contraction patterns. Our proposed method has shown some predictive power, however, further research is needed to improve the results for this important problem.<br/>}},
author = {{Simayijiang, Zhayida}},
isbn = {{978-91-7753-078-7}},
keywords = {{ToA, TDoA, Sensor Network Localization, EEG, CTG, machine learning.}},
language = {{eng}},
month = {{01}},
publisher = {{Lund University, Faculty of Science, Centre for Mathematical Sciences, Mathematics}},
school = {{Lund University}},
title = {{Applications of Signal Processing to Microphone Node Calibration and Medical Signal Classification}},
url = {{https://lup.lub.lu.se/search/files/19773265/thesis.pdf}},
year = {{2017}},
}