Hasil Pencarian  ::  Simpan CSV :: Kembali

"Penelitian ini mengembangkan sistem pendeteksi kebohongan berbasis kecerdasan buatan Machine learning untuk membantu petugas Imigrasi dalam mengevaluasi wawancara permohonan paspor warga negara Indonesia dengan pendekatan yang lebih sistematis melalui penggunaan teknologi computer vision guna menganalisis fitur-fitur yang ditampilkan oleh wajah manusia. Set data penelitian menggunakan 32 video wawancara (23 honest, 9 lie) yang diaugmentasi menjadi 41 video untuk mengatasi ketidakseimbangan kelas. Sistem menggunakan dlib dan OpenCV untuk mendeteksi 68 facial landmarks, mengekstrak tiga parameter perilaku wajah, yakni : bagian mata, bagian alis, dan bagian bibir. Analisis statistik comprehensive menghasilkan 66 fitur diskriminatif per video. Model CNN, CNN-LSTM, dan ANN dikembangkan untuk diujicobakan dalam penelitian ini. Model dilatih dengan 5 K-fold cross-validation sebanyak tiga kali dengan hasil evaluasi menunjukkan ANN mencapai test accuracy terbaik yaitu 74.4%, precision 79.3%, recall 74.4%, dan F1-score 73.4%. Analisis feature importance mengidentifikasi Mouth Asymmetry Std sebagai parameter paling diskriminatif dengan nilai sebesar 0.4114. Sistem kemudian diimplementasikan dalam web application dengan processing time sekitar lima menit per video, menyediakan confidence score dan final decision untuk analisis petugas. Kontribusi penelitian meliputi multi-feature facial analysis dan end-to-end system serta saran pengembangannya untuk aplikasi praktis wawancara paspor di kantor Imigrasi Indonesia yang diharapkan dapat menjadi solusi inovatif dalam mendukung pengambilan keputusan di instansi Imigrasi dan meningkatkan kualitas layanan publik terkait.

---

This research develops an artificial intelligence-based lie detection system using machine learning to assist Immigration officers in evaluating passport application interviews for Indonesian citizens through a more systematic approach utilizing computer vision technology to analyze features displayed by human faces. The research dataset uses 32 interview videos (23 honest, 9 lie) which were augmented to 41 videos to address class imbalance. The system employs dlib and OpenCV for detecting 68 facial landmarks, extracting three facial behavioral parameters: eye region, eyebrow region, and lip region. Comprehensive statistical analysis yields 66 discriminative features per video. CNN, CNN-LSTM, and ANN models were developed and tested in this research. Models were trained using 5 K-fold cross-validation performed three times, with evaluation results showing that ANN achieved the best test accuracy of 74.4%, precision of 79.3%, recall of 74.4%, and F1-score of 73.4%. Feature importance analysis identified Mouth Asymmetry Standard Deviation as the most discriminative parameter with a value of 0.4114. The system was then implemented as a web application with a processing time of approximately five minutes per video, providing confidence scores and final decisions for officer analysis. Research contributions include multi-feature facial analysis and end-to-end system development, along with recommendations for practical implementation in passport interview applications at Indonesian Immigration offices, which is expected to serve as an innovative solution in supporting decision-making at Immigration agencies and improving related public service quality."

"Selama masa pandemi COVID-19, antarmuka suara menggunakan KWS (keyword spotting) semakin sering digunakan pada berbagai sistem elektronik karena minimnya kontak fisik yang diperlukan antarmuka ini. Salah satu sistem yang dapat menggunakan KWS adalah sistem navigasi lift, di mana KWS pada sistem tersebut akan mengenali kata kunci terkait lantai yang ingin dituju pengguna. Dalam penelitian ini, model KWS untuk sistem navigasi lift dibuat menggunakan CNN (Convolutional Neural Network) dan CRNN (Convolutional Recurrent Neural Network) untuk mengenali enam kata kunci spesifik. Selama proses pembuatannya, berbagai hyperparameter CRNN terkait implementasi GRU, batch normalization, dropout layer, optimizer, kernel size, dan batch size diuji pengaruh variasinya terhadap performa CRNN. Dari pengujian tersebut, ditemukan bahwa CRNN menunjukkan performa paling baik ketika GRU yang digunakan bersifat bidirectional dengan dua layer dan 64 hidden unit, kernel size sebesar 3x3, optimizer Adams, batch size sebesar 163, serta penerapan batch normalization layer sebelum dropout layer. Model CRNN yang diperoleh dari kombinasi hyperparameter terbaik kemudian dibandingkan dengan model CNN untuk dievaluasi performa klasifikasinya saat dijalankan pada Raspberry Pi 4B. Berdasarkan hasil akurasi, persentase penggunaan RAM, dan latensi, model CNN menunjukkan performa yang lebih baik daripada CRNN.

---

During the COVID-19 pandemic, voice interfaces using KWS (keyword spotting) are increasingly being used in various electronic systems due to the lack of physical contact required for this interface. One system that can use KWS is an elevator navigation system, where the KWS on the system will recognize keywords related to the floor the user wants to go to. In this study, the KWS model for the elevator navigation system was created using CNN (Convolutional Neural Network) and CRNN (Convolutional Recurrent Neural Network) to identify six specific keywords. During the manufacturing process, various CRNN hyperparameters related to GRU implementation, batch normalization, dropout layer, optimizer, kernel size, and batch size were tested for the effect of their variations on CRNN performance. From these tests, it was found that CRNN showed the best performance when the GRU used bidirectional with two layers and 64 hidden units, kernel size of 3x3, Adams optimizer, batch size of 163, and batch normalization layer applied before dropout layer. The CRNN model obtained from the best combination of hyperparameters is then compared with the CNN model to evaluate its classification performance when run on the Raspberry Pi 4B. Based on the results of accuracy, percentage of RAM usage, and latency, CNN model shows better performance than CRNN."