Skip navigation


  • DSpace logo
  • Home
  • Collections
  • Researcher Profile
  • Explore by
    • Researcher Profile
  • VGU library
  • Help
  • User Guide
  • Sign on to:
    • My DSpace
    • Receive email
      updates
    • Edit Account details

VGU RESEARCH REPOSITORY


Please use this identifier to cite or link to this item: https://epub.vgu.edu.vn/handle/dlibvgu/586
DC FieldValueLanguage
dc.contributor.authorTran Chinhen_US
dc.date.accessioned2020-03-12T17:04:36Z-
dc.date.available2020-03-12T17:04:36Z-
dc.date.issued2017-
dc.identifier.urihttp://epub.vgu.edu.vn/handle/dlibvgu/586-
dc.description.abstractUltrasonic sensors have been first used to measure distance by calculating the time of flight of the echo. but they have been widely applied in many more fields such as robotics. manufacturing and automotive. Their main setbacks such as wide beam angle. slow propagation requires integration of either multiple ultrasonic modules or many different sensors technologies to improve their accuracy. In the past. researchers have been trying to combine ultrasonic sensors with machine learning, but back in the day the computational power was not enough to allow them to make high accuracies classification. With the today's breakthrough in machine learning algorithm and computational power. high accuracy classification with minimal ultrasonic sensor requirement will certainly leave an impact on many more fields. This thesis is set to solve the classification between human. automobile surface problem and blank space by using ultrasonic signals and machine learning algorithms at variable distance from 10 cm to 50 cm. In this project, human is regarded as an object having rough texture, which is represented by a cloth. while the automobile surface is regarded as having a smooth texture, which is represented by a smooth plastic board. The system differentiates the objects by analyzing both time and frequency domain of the signal using supervised machine learning methods. There are multiple machine learning schemes to be used to solve the problem of object classification such as logistic regression, neural network or support vector machines. Every method has its own advantages and disadvantages. This thesis shows performance comparison between the above methods working with different transducer setups. different datasets and the complexity of the dataset. In the given problem. with the same number of features, learning with time domain is almost always easier than frequency domain. Single channel setup with one receiver and one transmitter also performs better and more efficient than other setup in the given problem. Moreover, pre-processing data with principal component analysis yields better accuracy, and distance between transducers do not affect the system performance as long as the setup is consistent throughout training and testing phases.en_US
dc.language.isoenen_US
dc.publisherVietnamese-German Universityen_US
dc.rightsAttribution-NonCommercial 4.0 International*
dc.rights.urihttps://creativecommons.org/licenses/by-nc/4.0/*
dc.subjectObject classificationen_US
dc.subjectUltrasound systemen_US
dc.subjectMachine learning methodsen_US
dc.titleObject classification for ultrasound system using machine learning methodsen_US
dc.typeThesisen_US
item.fulltextWith Fulltext-
item.languageiso639-1other-
item.grantfulltextrestricted-
Appears in Collections:Electrical Engineering and Information Technology (EEIT)
Files in This Item:
File Description SizeFormat Existing users please Login
Object classification for ultrasound system using machine learning methods.pdf31.8 MBAdobe PDF
Show simple item record

Page view(s)

61
checked on Aug 23, 2025

Download(s)

29
checked on Aug 23, 2025

Google ScholarTM

Check


This item is licensed under a Creative Commons License Creative Commons

© Copyright 2020 by Vietnamese - German University Library.
Add: Ring road 4, Quarter 4, Thoi Hoa Ward, Ben Cat City, Binh Duong Province
Tel.:(0274) 222 0990. Ext.: 70206