KLASIFIKASI KEDIPAN MATA DALAM UPAYA MEMBANTU DISABILITAS UNTUK BERKOMUNIKASI MENGGUNAKAN RANDOM FOREST DAN XGBOOST

LUTHFI ARDI, Ir. Noor Akhmad Setiawan, S.T., M.T., Ph.D., IPM.

2021 | Tesis | MAGISTER TEKNOLOGI INFORMASI

Disabilitas adalah sebuah kekurangan (fisik atau mental), singkatnya suatu kondisi atau memiliki batasan untuk melakukan sesuatu. Dan ada beberapa kondisi yang membuat disabilitas sulit untuk berkomunikasi dengan orang lain. Saat ini banyak peneliti yang telah melakukan penelitan dengan tujuan membantu penyandang disabilitas untuk berkomunikasi dengan bantuan teknologi Brain Computer Interface, memanfaatkan salah satu Artifact dari perekaman Electroencephalograph (EEG) yaitu kedipan mata. Penelitian yang telah dilakukan kebanyakan hanya berfokus pada intention threshold dan puncak amplitude pada satu kedipan mata. Namun penelitian sebelumnya tidak membahas perbedaan jumlah sinyal kedipan berdasarkan jumlah kedipan mata yang didefinisikan sebagai satu kedipan, dua kedipan dan tiga kedipan. Penelitian ini menggunakan data primer yang diambil dengan menggunakan Muse Headband pada lima belas subjek. Data ini dapat digunakan sebagai dataset yang diklasifikasikan dengan metode bagging (Random Forest) dan boosting (XGBoost) dengan python, 80% data dialokasikan untuk training dan 20% untuk pengujuan. Data penelitian akan di uji hingga sepuluh kali pengujian, yang kemudian dirata-ratakan. Hasil klasifikasi jumlah kedipan mata menunjukkan nilai akurasi menggunakan Random Forest sebesar 97,8% dan hasil akurasi dengan metode XGBoost sebesar 98,5%. Hasil penelitian menunjukkan bahwa model eksperimen berhasil dan dapat digunakan sebagai referensi pembuatan aplikasi yang membantu orang disabilitas berkomunikasi membedakan jumlah kedipan mata.

Disability is a deficiency (physical or mental), in short, disabled people has a limit to do something. There are several conditions that make the disabilities have difficulty communicating with other people. Currently, many researchers that have helped people with disabilities by giving BCI technology, which uses Artifact Electroencephalograph (EEG) as a communication tool using blinking. Besides, research on eye blinks has only focused on intensions the threshold and peak amplitude by one blink regulation. However, previous research did not discuss the difference in signal blinking based on the number of eyes blinks defined as single blink, double blinks, and triple blinks. This study uses primary data taken using a muse headband on fifteen subjects. This data can be used as a dataset classified by bagging (Random Forest) and boosting (XGBoost) methods with python; 80% of the data is allocated for learning and 20% for testing. The classified data will be split up to ten times of testing, which are then averaged. The number of eye blinks classification results show that the accuracy value using Random Forest is 97.8%, and the accuracy results with the XGBoost method are 98.5%. The result has shown that the experimental model is successful and can be used as a reference for making applications that help people to communicate by differentiating the number of eye blinks. Disability is a deficiency (physical or mental), in short, disabled people has a limit to do something. There are several conditions that make the disabilities have difficulty communicating with other people. Currently, many researchers that have helped people with disabilities by giving BCI technology, which uses Artifact Electroencephalograph (EEG) as a communication tool using blinking. Besides, research on eye blinks has only focused on intensions the threshold and peak amplitude by one blink regulation. However, previous research did not discuss the difference in signal blinking based on the number of eyes blinks defined as single blink, double blinks, and triple blinks. This study uses primary data taken using a muse headband on fifteen subjects. This data can be used as a dataset classified by bagging (Random Forest) and boosting (XGBoost) methods with python; 80% of the data is allocated for learning and 20% for testing. The classified data will be split up to ten times of testing, which are then averaged. The number of eye blinks classification results show that the accuracy value using Random Forest is 97.8%, and the accuracy results with the XGBoost method are 98.5%. The result has shown that the experimental model is successful and can be used as a reference for making applications that help people to communicate by differentiating the number of eye blinks.

Kata Kunci : BCI, EEG, Eye Blink, Disability, Random Forest, XGBoost