Emotion Recognition Algorithms and Feedback Mechanisms
The system uses a two-layer emotion detection model:
- keyword matching:: 200+ predefined emotion words (e.g. "happy" corresponds to a joy score of +2)
- speech analysis: Recognize pitch changes through spectral features (percentage of high frequency components reflecting excitability)
The 6 basic emotional feedbacks currently supported:
- Happiness: increase the frequency of emoji emoji and tone of voice (~) usage
- Anger: extended response interval + pacifying phrases
- Sadness: turn down the pitch of the synthesized voice + add a comforting statement
- Confusion: automatically follow up with clarifying questions + provide examples
- Fatigue: simplify response structure + proactively suggest breaks
- Neutral: maintain a standard professional tone
This answer comes from the articleXiaozhi MCP Client: a cross-platform AI assistant supporting voice and text interactionThe
May not be reproduced without permission:AI productivity tools " How does the Mood Mode in the Kotaku client recognize user emotions? What emotional state feedback is supported?
