HENRY ARDIAN IRIANTA (2022) REKAYASA TINY MACHINE LEARNING MULTI KENDALI DENGAN METODE CONVOLUTIONAL NEURAL NETWORK PADA EXO GLOVE UNTUK TERAPI PASCA STROKE. S1 thesis, Universitas Muhammadiyah Yogyakarta.

	Text (Halaman Judul) Halaman Judul.pdf Download (286kB)
	Text (Lembar Pengesahan) Lembar Pengesahan.pdf Restricted to Registered users only Download (2MB)
	Text (Abstrak) Abstrak.pdf Restricted to Registered users only Download (29kB)
	Text (Bab I) Bab I.pdf Download (149kB)
	Text (Bab II) Bab II.pdf Restricted to Registered users only Download (1MB)
	Text (Bab III) Bab III.pdf Restricted to Registered users only Download (1MB)
	Text (Bab IV) Bab IV.pdf Restricted to Registered users only Download (1MB)
	Text (Bab V) Bab V.pdf Restricted to Registered users only Download (76kB)
	Text (Daftar Pustaka) Daftar Pustaka.pdf Restricted to Registered users only Download (153kB)
	Text (Lampiran) Lampiran.pdf Restricted to Registered users only Download (202kB)
	Text (Naskah Publikasi) Naskah Publikasi.pdf Restricted to Registered users only Download (1MB)
	Text (Full Text) Full Text.pdf Restricted to Repository staff only Download (4MB)

Abstract

Underactuated Exo glove, have received an alternative robot rehabilitation and hand grasp assistance for improving activity daily living (ADL) on particular severe, like neurodegenerative and musculoskeletal diseases. These are hemaperese and dysarthria after post stroke. Despite the significant progress in this field, the control system that hase been used mostly uses a complex controller, is bulky, and expensive. The purpose of this study is to develop a low-cost wirelessly multi-controller, with 2 operations, using micro speech recognition including disartric and nondisartic, and with a momentary button, under 300k IDR. The speech recognition interface using, INMP441 and identified using a Convolutional Neural network (CNN) and embedded in ESP-32. Voice data will be extracted and converted to spectrogram feuture using Short Time Fourier Transform (STFT) method and fed into CNN. This method has proven be useful in micro speech recognition on both speech scenarios with level accuracy above 90%. Realtime inference performance on ESP-32 using hand prostethic, with 3 level household noise intensity with 24db,42db, and 62db, has respectively resulted from 95%, 85%, 50% Accuracy. Wireless connectivity success rate with both types of controllers respectively around 0.2 - 0.5 ms.

Dosen Pembimbing:	BAMBANG RIYANTA, DR., S.T., M.T. and BERLI PARIPURNA KAMIEL, IR., S.T., M.ENG.SC., PH.D. | NIDN0524017101, NIDN0502037401
Item Type:	Thesis (S1)
Uncontrolled Keywords:	EXO GLOVE, MICRO SPEECH RECOGNITION, TINYML, STFT, CNN, FOSS
Divisions:	Fakultas Teknik > S1 Teknik Mesin
Depositing User:	M. Erdiansyah
Date Deposited:	20 May 2022 02:10
Last Modified:	20 May 2022 02:10
URI:	https://etd.umy.ac.id/id/eprint/31335

Actions (login required)

View Item