El hareketlerini tanıma için makine öğrenmesi algoritmalarının uygulanması
Loading...

Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Hareket tanıma, insan-bilgisayar etkileşimi (HCI) için son derece önemlidir. Bir el hareketi tanıma sistemi, sözlü olmayan iletişimin doğal, yenilikçi ve modern bir yolunu sağlar. İnsan-bilgisayar etkileşimlerinde geniş bir uygulama alanına sahiptir. El hareketlerinin bilgisayarla tanınması, tıbbi sistemler, insan-bilgisayar etkileşimi gibi birçok uygulamada yaygın olarak kullanılmaktadır, çünkü el hareketi tanıma, insanlara doğal ve sezgisel bir bilgisayar ara yüzü sağlar. Bu çalışmada, kullanıcının artık kas hareketlerini; el protezinin açık / kapalı el, el bileğini döndürmek gibi belirli hareketlerini haritalanması amaçlanmıştır. Bu problemleri çözmek için öncelikle, hangi özelliklerin gerekli olduğuna karar vermek için, bazı sütunları varyasyonları halinde çıkararak deneyler gerçekleştirildi. Oluşturulan veri kümesi ile yapay sinir ağları algoritmalarından birçoğu ile deneyler yapılmış, Naive Bayes, BayesNet, Multilayer Perceptron, Bagging, Hoeffding Tree and Random Forest olan en başarılı algoritmalar arasından Random Forest seçilmiştir.
Motion recognition is extremely important for human-computer interaction (HCl). A hand gesture recognition system contributes an organic, creative and brought up to date version of non-verbal communication. HCI has a wide range of applications such as computer recognition of gestures, medical systems, human-robot interaction because gesture recognition make available to people with a characteristic and instinctive computer interface. The aim of the thesis is to map user residual muscle gestures to certain actions of a prosthetic such as open/close hand or rotate the wrist. For this propose, firstly, in order to decide which features are necessary, experiments were performed by removing some features in combination. With created datasets, experiments have been done with many of the artificial neural network algorithms. Random Forest was chosen among the most successful algorithms which are Naive Bayes, BayesNet, Multilayer Perceptron, Bagging, Hoeffding Tree and Random Forest.
Motion recognition is extremely important for human-computer interaction (HCl). A hand gesture recognition system contributes an organic, creative and brought up to date version of non-verbal communication. HCI has a wide range of applications such as computer recognition of gestures, medical systems, human-robot interaction because gesture recognition make available to people with a characteristic and instinctive computer interface. The aim of the thesis is to map user residual muscle gestures to certain actions of a prosthetic such as open/close hand or rotate the wrist. For this propose, firstly, in order to decide which features are necessary, experiments were performed by removing some features in combination. With created datasets, experiments have been done with many of the artificial neural network algorithms. Random Forest was chosen among the most successful algorithms which are Naive Bayes, BayesNet, Multilayer Perceptron, Bagging, Hoeffding Tree and Random Forest.
Description
Keywords
Algorithms, Human-Machine Relationship, Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Algoritmalar, İnsan-Teknoloji Ilişkisi, Machine Learning Methods, Human-Technology Relationship, İnsan-Makine Ilişkisi, İnsan-Makine Sistemi, Artificial Neural Networks, Man-Machine System, Hand Recognition, Computer Engineering and Computer Science and Control, El Tanıma, Yapay Sinir Ağları, Makine Öğrenmesi Yöntemleri
Turkish CoHE Thesis Center URL
Fields of Science
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
51
