Prakoso, Khrisna Shane Budy and Barlian Henryranu Prasetio, S.T., M.T., Ph.D. (2025) Implementasi Speech Recognition berbasis Raspberry Pi 5 pada sistem Smart-Home menggunakan Algoritma Gated Recurrent Unit (GRU). Sarjana thesis, Universitas Brawijaya.
Abstract
Sistem smart home terus berkembang untuk memberikan kemudahan dalam kehidupan sehari-hari, dengan pengenalan suara menjadi elemen penting. Namun, tantangan seperti gangguan noise, variasi aksen, dan intonasi pengguna sering memengaruhi akurasi sistem. Penelitian ini bertujuan mengembangkan sistem pengenalan suara berbasis Raspberry Pi 5 dengan algoritma Gated Recurrent Unit (GRU) untuk meningkatkan akurasi dan performa dalam mengenali perintah suara pada sistem smart home. Metodologi penelitian mencakup perekaman dataset suara dengan variasi perintah, ekstraksi fitur menggunakan Mel-Frequency Cepstral Coefficients (MFCC), pelatihan model GRU dengan optimasi parameter, dan pengujian real-time dalam kondisi lingkungan nyata. Evaluasi dilakukan untuk mengukur akurasi dan robustitas sistem terhadap noise dan variasi pola suara pengguna. Hasil pengujian menunjukkan sistem mampu mengenali perintah suara dengan akurasi rata-rata sebesar 88.75% pada kondisi noise rendah hingga sedang. Sistem berhasil mengenali berbagai perintah, seperti mengontrol perangkat rumah tangga dan membuka pintu. Tantangan yang dihadapi meliputi seperti prediksi salah akibat pola suara mirip antar-perintah, sensitivitas terhadap noise tinggi, serta variasi aksen dan intonasi yang kurang terwakili dalam dataset. Respons "tidak yakin" menandakan perlunya pengayaan dataset dan penerapan teknik preprocessing seperti noise filtering. Penelitian ini berkontribusi dalam pengembangan sistem pengenalan suara berbasis perangkat kecil dengan daya rendah untuk smart home. Future work mencakup eksplorasi algoritma hybrid, seperti GRU-LSTM atau transformer-based models, serta pengayaan dataset untuk meningkatkan performa sistem dalam berbagai skenario. Dengan langkah ini, sistem dapat menjadi solusi yang lebih fleksibel dan andal bagi teknologi smart home yang terus berkembang.
English Abstract
Smart home systems continue to evolve, offering convenience in daily life with speech recognition as a key element. However, challenges such as noise interference, accent variations, and user intonation often impact system accuracy. This research aims to develop a speech recognition system based on Raspberry Pi 5, integrating the Gated Recurrent Unit (GRU) algorithm to enhance accuracy and performance in recognizing voice commands within the smart home ecosystem. The research methodology includes recording a voice dataset with command variations, feature extraction using Mel-Frequency Cepstral Coefficients (MFCC), training the GRU model with parameter optimization, and conducting real-time testing in real-world environments. Evaluations were performed to measure the system's accuracy and robustness against noise and variations in user speech patterns. The results indicate that the system can recognize voice commands with an average accuracy of 88.75% under low to moderate noise conditions. The system successfully identifies various commands, such as controlling household devices and unlocking doors. However, challenges remain, including incorrect predictions due to similar voice patterns between commands, high sensitivity to noise, and unrepresented variations in accents and intonations within the dataset. The "uncertain" response highlights the need for dataset enrichment and preprocessing techniques like noise filtering. This research contributes to the development of low-power, compact speech recognition systems for smart homes. Future work involves exploring hybrid algorithms, such as GRU-LSTM or transformer-based models, and expanding the dataset to enhance system performance across diverse scenarios. These efforts aim to create a more flexible and reliable solution for the continuously evolving smart home technology.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | 052515 |
| Uncontrolled Keywords: | Pengenalan suara, Raspberry Pi 5, Gated Recurrent Unit, Smart home, Noise, MFCC. |
| Divisions: | Fakultas Ilmu Komputer > Teknik Komputer |
| Depositing User: | S Sucipto |
| Date Deposited: | 20 Feb 2025 01:51 |
| Last Modified: | 20 Feb 2025 01:51 |
| URI: | http://repository.ub.ac.id/id/eprint/237174 |
![[thumbnail of DALAM MASA EMBARGO]](http://repository.ub.ac.id/style/images/fileicons/text.png) |
Text (DALAM MASA EMBARGO)
Khrisna Shane Budy Prakoso.pdf Restricted to Registered users only Download (4MB) |
Actions (login required)
 |
View Item |