research

Full content with pdf file

Full content with pdf file

Full content with pdf file

📱 TalkingPicture – AI-Powered Image Description App

Helping visually impaired “see” the world through spoken words

TalkingPicture is a mobile application designed to support people with visual impairments in perceiving their surroundings. By using advanced AI technology, the app analyzes captured images and provides natural-sounding audio descriptions in either Vietnamese or English. With a fully accessible and intuitive interface, users can interact entirely through touch, sound cues, and vibration — no visual navigation required.

🔍 Key Capabilities

1. Tap anywhere on the screen to capture a photo and receive an instant spoken description
2. Dual-language support: Vietnamese & English
3. Two description modes: concise or detailed
4. Automatic saving and playback of the 10 most recent images
5. Gesture-based controls optimized for accessibility
6. Voice feedback, audio instructions & haptic cues for every action

🚀 How to Use TalkingPicture

1. Launch the App
2. Open TalkingPicture — voice guidance will walk you through all basic gestures (tap, swipe, hold). The app starts directly in Camera mode.
3. Camera Screen
4. Tap anywhere → take a photo
5. Phone vibrates + sound effect confirms the shot
6. Wait a moment → the app speaks the AI-generated image description
7. Hold the screen → replay the last description
8. Tap again → take the next photo
9. History Screen
10. Swipe up/down to browse saved images
11. Each image automatically plays its stored description
12. Tap → replay the current description
13. Settings Screen
14. Swipe left/right to switch between categories
15. Tap or swipe up/down to adjust language, speech speed, and description length
16. Every adjustment is confirmed by a voice cue

✨ Quick Tip: Remember 3 simple actions — Tap to capture • Hold to replay • Swipe to navigate

🌟 Our Mission

TalkingPicture empowers visually impaired users to experience the world with greater independence — transforming every image into a meaningful story told through sound.

(Illustration Video: https://youtube.com/shorts/lNiblK_f6sk?feature=share)

Dang, MN., Cao, TH., Lam, TD., Le, MH., Le, DN., Dinh, SD. (2026). Parallel Corpus Construction for Chinese and Vietnamese in Historical Texts.

1. In: Nguyen, N.T., et al. Computational Intelligence in Engineering Science. ICCIES 2025. Communications in Computer and Information Science, vol 2585. Springer, Cham.
2. Abstract: This study builds a Vietnamese–ancient Chinese parallel corpus from the Đại Nam Thực Lục Tiền Biên and introduces a phrase-matching algorithm that surpasses BertAlign, offering a promising approach for historical text alignment.

Thi, L., Dien, D.S., Tiep, N.V. (2026). Scene Sino-Nom Text Optical Character Recognition in Vietnam: A Novel Approach Combining Deep Learning and Linguistic Knowledge

1. In: Nguyen, N.T., et al. Computational Intelligence in Engineering Science. ICCIES 2025. Communications in Computer and Information Science, vol 2587. Springer, Cham.
2. Abstract: This paper presents Vietnam’s first outdoor Sino-Nom character recognition system, achieving 80% accuracy through deep learning and linguistic knowledge to support the preservation of cultural heritage.