Prototipe Sistem Deteksi Perintah Dan Klasifikasi Kelompok Usia Berbasis Suara Pada Perangkat Tiny Machine Learning (TINYML)

Dewandra, Abadila Barasmara Bias (2024) Prototipe Sistem Deteksi Perintah Dan Klasifikasi Kelompok Usia Berbasis Suara Pada Perangkat Tiny Machine Learning (TINYML). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5027201052-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 July 2026. Download (3MB) | Request a copy

Abstract

Suara merupakan identitas seseorang yang memiliki ciri-ciri unik. Seiring bertambahnya usia, suara mengalami perubahan dan menunjukkan karakteristik khusus pada berbagai tahapan kehidupan. Meskipun sistem klasifikasi suara telah banyak diintegrasikan dalam berbagai perangkat pintar untuk melaksanakan tugas tertentu, masih terdapat beberapa kelemahan, terutama dalam bidang keamanan, di mana sistem cenderung tidak memperhatikan faktor usia pengguna dan hanya menerima suara perintah tertentu. Penelitian ini bertujuan untuk mengembangkan prototipe sistem suara yang mampu menerima perintah suara sekaligus mengklasifikasikan usia pengguna. Pengembangan menggunakan model One-Dimensional Convolutional Neural Network (1D CNN) dengan ekstraksi fitur menggunakan Mel Frequency Cepstral Coefficient (MFCC) dan membandingkannya dengan model klasifikasi usia sebelumnya di berbagai dataset. Model diintegrasikan ke perangkat Arduino Nano 33 BLE Sense Lite dan dikembangkan menggunakan platform Edge Impulse. Pengujian dilakukan dengan variasi skenario kelompok usia dan perintah 'buka' dan 'tutup'. Skenario terbaik akan dipilih berdasarkan jumlah label terbanyak dengan akurasi terbaik. Sistem pada skenario terbaik memiliki akurasi terkuantisasi sebesar 89.59%. Setelah proses optimasi, akurasi meningkat menjadi 97.91%. Model akan dikembangkan menjadi gerendel pintar yang menggerakkan kunci sesuai jenjang usia tertentu. Hasil ini menunjukkan potensi sistem untuk diterapkan dalam konteks keamanan dan autentikasi suara dengan sumber daya terbatas
====================================================================================================================================
Voice is the identity of a person who has unique characteristics. As we age, the voice changes and exhibits special characteristics at different stages of life. Although voice classification systems have been widely integrated in various smart devices to carry out certain tasks, there are still some weaknesses, especially in the field of security, where the systems tend not to pay attention to the user's age and only accept certain voice commands. This research aims to develop a voice system prototype that is capable of receiving voice commands while classifying the user's age. The development uses a One-Dimensional Convolutional Neural Network (1D CNN) model with feature extraction using the Mel Frequency Cepstral Coefficient (MFCC) and compares it with previous age classification models in various datasets. The model is integrated into an Arduino Nano 33 BLE Sense Lite device and developed using the Edge Impulse platform. Testing was carried out with a variety of age group scenarios and 'open' and 'close' commands. The best scenario will be selected based on the largest number of labels with the best accuracy. The system in the best scenario has a quantized accuracy of 89.59%. After the optimization process, the accuracy increased to 97.91%. The model will be developed into a smart deadbolt that moves the key according to certain age levels. These results demonstrate the system's potential for application in resource-limited voice security and authentication contexts.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Perintah, Klasifikasi Usia Anak, One-Dimensional Convolutional Neural Network (1D CNN), Suara, TinyML, Children Age Classification; Command Detection, One-Dimensional Convolutional Neural Network (1D CNN), TinyML, Voice.
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Abadila Barasmara Bias Dewandra
Date Deposited:	05 Feb 2024 06:16
Last Modified:	05 Feb 2024 06:16
URI:	http://repository.its.ac.id/id/eprint/106094

Actions (login required)

View Item