Pengenalan Fonik Menggunakan Metode Deep Learning

Yulianto, Nayya Kamila Putri (2025) Pengenalan Fonik Menggunakan Metode Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025211183-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Pendidikan fonik merupakan dasar penting dalam kemampuan membaca dan mengeja karena mengajarkan hubungan antara bunyi dan huruf. Fonik terbukti meningkatkan kemampuan anak dalam memahami dan memperlancar bacaan dengan membantu mereka mengidentifikasi pola bunyi huruf. Namun, di lingkungan multibahasa seperti Indonesia, penerapan fonik menjadi tantangan akibat keragaman dialek lokal yang memengaruhi pelafalan huruf, khususnya pada penutur non-native Bahasa Inggris. Tugas Akhir ini menggunakan teknologi pengenalan suara berbasis deep learning untuk mengidentifikasi fonik. Sebanyak 986 audio dikumpulkan dari 38 penutur dan diproses melalui augmentasi. Fitur suara diekstraksi dengan Mel Frequency Cepstrum Coefficients (MFCC) dan Power-Normalized Cepstral Coefficients (PNCC), kemudian dilatih menggunakan Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), dan Transformer. Evaluasi dilakukan menggunakan nilai akurasi dan F1 Score. Hasil terbaik diperoleh dari model RNN-GRU dan ekstraksi fitur PNCC yang mencapai akurasi 94.59% dan F1 Score 0.946. Hasil ini melampaui penelitian sebelumnya yang menggunakan Support Vector Machine (SVM) dan ekstraksi fitur MFCC Librosa dengan nilai akurasi 91.95% dan F1 Score 0.9242.
==================================================================================================================
Phonics education is a fundamental foundation in developing reading and spelling skills by teaching the relationship between sounds and letters. It has been shown to enhance children's reading fluency and comprehension by helping them recognize sound patterns in words. However, in multilingual environments such as Indonesia, implementing phonics presents challenges due to diverse local dialects that affect pronunciation, particularly among non-native English speakers. This Final Project employs speech recognition technology based on deep learning to identify phonics. A total of 986 audio recordings were collected from 38 speakers and processed through data augmentation methods. Audio features were extracted using Mel-Frequency Cepstral Coefficients (MFCC) and Power-Normalized Cepstral Coefficients (PNCC), then trained using Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Transformer. The models were evaluated using accuracy and F1 Score metrics. The best performance was achieved by the RNN-GRU model with PNCC features, reaching an accuracy of 94.59% and an F1 Score of 0.946. This result outperformed the previous study that employed Support Vector Machine (SVM) with MFCC Librosa features, which obtained an accuracy of 91.95% and an F1 Score of 0.9242.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Pengenalan fonik, deep learning, dialek Indonesia, phonics recognition, Indonesian dialect
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Nayya Kamila Putri Yulianto
Date Deposited:	29 Jul 2025 03:16
Last Modified:	29 Jul 2025 03:16
URI:	http://repository.its.ac.id/id/eprint/122428

Actions (login required)

View Item