🧠 AI Ed Wiki

📄 Research Article

AI-Guided Learning: Research on Knowledge and Skill Acquisition Support Methods Using Deep Learning Audio-Video Processing Techniques

● high · 2026-08-12personalized-learning, language-learning, feedback-loop, self-regulated-learning, multimodal, student-modeling, learning-analytics, lifelong-learning

Synthesis: This dissertation develops an AI-guided learning framework that supports three interconnected stages — Consume, Understand, and Imitate — with three deep-learning systems for audio/video learning. AIxSpeed adapts audio playback speed using speech-recognition confidence; FastPerson produces multimodal video summaries; and Profy supports pronunciation practice from largely unannotated speech. Evaluations show efficiency gains (up to 1.30x playback, 53% less viewing time) with no loss in learning outcomes, and improved pronunciation intelligibility.

Key Findings

1. Consume — adaptive audio speed. AIxSpeed dynamically adjusts playback speed at the phoneme level using speech-recognition confidence as a proxy for listening difficulty, achieving average playback factors of 1.30x (LibriSpeech) and 1.29x (UME-ERJ) with higher mean opinion scores than matched constant-speed playback.

2. Understand — multimodal summaries. FastPerson generates video summaries preserving visual and auditory information and lets learners switch between summarized and full versions by chapter, reducing viewing time by 53% with no statistically significant difference in quiz scores.

3. Imitate — proficiency feedback. Profy learns proficiency from largely unannotated speech, visualizes classifier-relevant regions and model-derived acoustic distances, and produced an observed improvement in pronunciation intelligibility with non-overlapping pre/post confidence intervals.

Implications

The framework illustrates how Personalized Learning can address the time cost of sequential audio/video content and the lack of scalable feedback for imitation-based skill acquisition, notably in Language Learning pronunciation. The systems exemplify a Feedback Loop design where model confidence and learner behavior jointly drive adaptation, and they connect to Multimodal and Student Modeling research.

By preserving learner access to the original material while adding AI assistance, the work aligns with Self Regulated Learning and Lifelong Learning goals — supporting efficient consumption and repeated practice without removing learner agency. The three-stage Consume–Understand–Imitate framing offers a reusable model for AI-guided media learning.

Connected Concepts

  • Personalized Learning
  • Language Learning
  • Feedback Loop
  • Self Regulated Learning
  • Multimodal
  • Student Modeling
  • Learning Analytics
  • Lifelong Learning

    • Connected Articles

  • AI Assisted Learning Modes Eeg — AI-assisted learning modes (EEG)
  • Lecture Review — Lecture review with AI
  • AI Generated Instructional Videos Computing Ed — AI-generated instructional videos
  • Not A Universal Benefit Examining The Differential Effects Of Emotional AI On L2 — Emotional AI on L2 learning
  • AI Adult Learning Design — AI in adult learning design

    • Citation

    Kawamura, K. (2026). AI-guided learning: Research on knowledge and skill acquisition support methods using deep learning audio-video processing techniques. arXiv:2608.08990. cs.HC. Doctoral dissertation, University of Tokyo.