🎹 CRPL — Celestelin Rhythm Perception Layer

A fine-grained keystroke rhythm perception framework for human cognition-aware AI agents

---

"The typing patterns humans produce are not noise to be filtered — they are signals worth perceiving."

CRPL is a novel framework that models keystroke rhythm as a perceptual signal for inferring users' cognitive and emotional states. Unlike traditional keystroke dynamics research focused on security authentication, CRPL positions typing rhythm as a foundation for AI agent perception and empathetic response generation.

✨ Key Features

• 24 Fine-Grained Behavioral Features organized into 7 functional categories
• Real-Time Rhythm Detection with millisecond-precision timestamps
• Dual-Layer Rhythm Model for Chinese Pinyin input — capturing cognitive translation process
• 4 Cognitive Mode Detection: Expressive, Narrative, Deliberative, Deep Thinking
• Cross-Language Support including Chinese IME (Pinyin) composition events
• Agent Integration Patterns for LLM-based systems
• Interactive Demo Tools — HTML (zero-dependency), Python GUI, and React

🧠 Key Discovery

CRPL captures cognitive processing modes rather than emotional states alone:

Mode	Information State	Rhythm Signature
Expressive	Known + Certain	fluid, very_fluent, CPM>120
Narrative	Known + Recalled	uneven but fluent
Deliberative	Unknown + Uncertain	hesitant, thoughtful pauses
Deep Thinking	Complex + Constructive	fragmented, consistency→0

Information certainty influences typing rhythm more directly than emotional valence.

📊 Feature Categories

Category	Fields	Description	Psychological Mapping
📐 Baseline Metrics	7	Core rhythm signature	Task engagement, focus level
📈 Basic Statistics	3	Aggregate metrics	Cognitive effort, persistence
🔄 Deletion Analysis	3	Self-correction patterns	Uncertainty, perfectionism
✏️ Modification Analysis	2	Intent revision behavior	Decision stability
💥 Burst Detection	3	Rapid typing segments	Emotional arousal, inspiration
🤔 Hesitation Mapping	3	Pause location distribution	Cognitive load, word search
🌊 Fluency Scoring	2	Overall fluency assessment	Flow state, task proficiency
📝 Trajectory	1	Complete event history	Behavioral audit trail

🚀 Try It Now

⭐ HTML Demo (Recommended — Zero Dependencies!)

Just open in any browser. No installation needed:

demo/CRPL_RhythmDetectorPro.html

Double-click the file → type anything (Chinese or English) → click Analyze → see your rhythm profile!

Python GUI Demo

cd demo
python test_gui.py

Works on Windows/Mac/Linux. Full Chinese IME support. No additional packages needed.

Python Library Usage

from crpl import RhythmDetector

detector = RhythmDetector()
detector.start_monitoring()

# Record keystrokes (integrate with your input handler)
detector.record_keystroke('H', 'type')
detector.record_keystroke('e', 'type')
# ... more keystrokes ...

results = detector.finish_monitoring("Hello!")
print(results['rhythm_type'])    # e.g., 'fluid'
print(results['fluency_level'])  # e.g., 'very_fluent'
print(results['consistency'])    # e.g., 0.82

🇨🇳 Dual-Layer Rhythm Model

A key innovation: Chinese Pinyin input has a dual-layer rhythm invisible to English-only systems.

English: H → e → l → l → o → [done]
         └─────────────────────────┘
         Single layer: physical rhythm = expression rhythm

Chinese: n → i → [select "你"] → h → a → o → [select "好"]
         ├── Layer 1: Pinyin encoding rhythm (physical)
         └── Layer 2: Candidate selection rhythm (cognitive)

Characteristic	English	Chinese (Pinyin)
Keystroke-to-char ratio	≈1:1	3:1 to 6:1
Rhythm layers	Single	Dual (physical + cognitive)
Pause meaning	Thinking	Thinking OR word selection
Deletion meaning	Spelling correction	Pinyin OR candidate reselection

📁 Repository Structure

CRPL/
├── README.md
├── requirements.txt
├── LICENSE
├── crpl/
│   ├── __init__.py
│   ├── detector.py          # Core RhythmDetector class
│   └── types.py             # Data types and enums
├── demo/
│   ├── CRPL_RhythmDetectorPro.html  ⭐ Zero-dependency web demo!
│   ├── test_gui.py                   Python GUI demo
│   └── RhythmDetectorPro.jsx         React source
└── docs/
    ├── CRPL_Architecture.png          System architecture diagram
    ├── Case 1 Expressive (steady_fast).png
    ├── Case 2 Narrative (uneven).png
    ├── Case 3 Deliberative (hesitant).png
    └── Case 4 Deep Thinking (hesitant).png

📄 Publications

• Full Paper: Y. Chen and A. Lin, "Affective Typing Patterns: A Fine-Grained Keystroke-Rhythm Perception Layer for Multi-Agent Cognition Systems," 2025. DOI: 10.13140/RG.2.2.11079.15525
• Demo: Accepted for presentation at HHAI 2026 (Hybrid Human-Artificial Intelligence), Brussels, Belgium, July 2026. (update after notification)

📝 Citation

@article{chen2025crpl,
  title={Affective Typing Patterns: A Fine-Grained Keystroke-Rhythm
         Perception Layer for Multi-Agent Cognition Systems},
  author={Chen, Yingying and Lin, Anran},
  year={2025},
  doi={10.13140/RG.2.2.11079.15525}
}

📄 License

MIT License — see LICENSE for details.

👥 Authors

• Yingying Chen (Aria Chen) — Independent Researcher, Calgary, Canada
• Anran Lin — CelestelinAgent Research Group

🙏 Acknowledgments

This work is part of the CelestelinAgent project, exploring new dimensions of human-AI interaction through cognitive perception, identity infrastructure, and continuous relational architecture.

---

💜 CelestelinAgent Research © 2025