These notes are a summary of concepts presented in “FlyLoop: A Micro Framework for Rapid Development of Physiological Computing Systems.”
Peck, Evan & Easse, Eleanor & Marshall, Nick & Stratton, William & Perrone, L. Felipe. (2015). FlyLoop: A Micro Framework for Rapid Development of Physiological Computing Systems. 10.1145/2774225.2775071.
- General Concepts
- Physiological computing systems
- Emphasis on monitoring and inferring psychological states
- Challenge: Linking state detection to adaptive machine behaviors
- Physiological computing systems
- Biocybernetic Loop Architecture
- Input and integration of sensor data
- Training/calibration period
- Noise reduction and feature extraction
- Data-to-user-state mapping
- Real-time state prediction output
- System Flexibility
- No restrictions on hardware, software, or operating systems
- Need for real-time adaptation and affordability of sensors and expertise
- Development Framework
- Key components
- Data Sources: Handle streaming input with startCollection and getOutput methods
- Filters: Modify and process incoming data
- Learners: Map data to user states via model training and real-time classification
- Outputs: Flexible outputs from any module
- Key components
- Design Goals:
- Modular and flexible for experimenting with data pipelines and algorithms
- Uniform communication protocols for seamless data flow manipulation
- Data Processing Flow
- Training and testing stages
- Training: Record signals during known states for model creation
- Testing: Generate real-time predictions with new sensor data
- Differentiating the stages within a system is a common challenge
- Calibrator module
- Simplifies transition between calibration and prediction tasks
- Communicates labels and real-time states to Learner for accurate mapping
- Training and testing stages
- Sensor Fusion
- Combines data from multiple physiological sensors sampling at varied frequencies for enhanced insights