Pengembangan Aplikasi Audio Relaksasi Mobile dengan Sistem Rekomendasi Berbasis Rekognisi Ekspresi Wajah (Studi Kasus: Mahasiswa Generasi Z)

Kesehatan mental mahasiswa merupakan isu krusial di Indonesia, khususnya bagi Generasi Z yang rentan terhadap stres akademik dan tekanan digital. Terapi audio relaksasi terbukti efektif menurunkan stres, namun implementasi yang bersifat personal dan adaptif masih terbatas. Penelitian ini bertujuan mengembangkan aplikasi mobile "Teduhin" yang mampu mendeteksi ekspresi wajah pengguna dan memberikan rekomendasi audio relaksasi secara otomatis berdasarkan emosi yang terdeteksi secara real-time. Sistem ini mengintegrasikan model machine learning berbasis arsitektur MobileNetV2 yang dikonversi ke format TensorFlow Lite untuk deteksi emosi wajah secara on-device guna menjamin privasi pengguna. Model dilatih menggunakan teknik transfer learning dan fine-tuning, menghasilkan nilai akurasi akhir sebesar 68,00% dengan puncak akurasi validasi mencapai 68,62%. Selain itu, aplikasi dilengkapi asisten percakapan berbasis AI untuk dukungan relaksasi tambahan. Metode pengembangan divalidasi melalui uji pakar informatika dan psikologi, sementara pengalaman pengguna dievaluasi menggunakan kuesioner terhadap responden mahasiswa Generasi Z. Hasil pengujian menunjukkan bahwa sistem berhasil menjalankan deteksi emosi dan pemutaran audio secara stabil tanpa ketergantungan pada koneksi internet. Evaluasi usability menghasilkan skor rata-rata 61,14 (Marginal High), dengan skor efektivitas audio relaksasi sebesar 3,91/5,0 dan kesesuaian fitur terhadap kondisi stres sebesar 4,18/5,0. Penelitian ini memberikan kontribusi pada pengembangan intervensi kesehatan mental non-farmakologis yang personal, adaptif, dan mudah diakses bagi mahasiswa.

===============================================================================================

Students’ mental health is a crucial issue in Indonesia, especially for Generation Z, who are vulnerable to academic stress and digital pressure. Relaxation audio therapy has been proven effective in reducing stress, but personalized and adaptive implementations are still limited. This research aims to develop the “Teduhin” mobile application that can detect users’ facial expressions and automatically provide relaxation audio recommendations based on the detected emotions in real-time. This system integrates a machine learning model based on the MobileNetV2 architecture converted to TensorFlow Lite format for on-device facial emotion detection to ensure user privacy. The model was trained using transfer learning and fine-tuning techniques, resulting in a final accuracy of 68.00%, with a peak validation accuracy of 68.62%. Additionally, the app features an AI-based conversational assistant for additional relaxation support. The development method was validated through expert testing in informatics and psychology, while user experience was evaluated using a questionnaire with Generation Z student respondents. Test results showed that the system successfully performed emotion detection and audio playback stably without relying on an internet connection. The usability evaluation yielded an average score of 61.14 (Marginal High), with a relaxation audio effectiveness score of 3.91/5.0 and a feature suitability score of 4.18/5.0 for stressful conditions. This research contributes to the development of personalized, adaptive, and accessible non-pharmacological mental health interventions for college students.