Keynote: Using EEG and Machine Learning to Perform Lie Detection
summary
14 EEG channels recorded with an Emotiv EPOC headset at 128 Hz sampling produce labeled truth/lie segments (manual markers) that are fed into Azure Machine Learning to train a classifier. A per-person workflow (husband, manager) yields a per-person classifier deployed as a REST endpoint and a simple Windows 10 app, showing up to ~93% accuracy on held-out test folds. Limitations: models are person-specific, need more subjects, better CSV automation and deeper models.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:16What is EEG
- 2:58Demo
- 13:01How I Got This Idea
- 14:42Experiment Procedure
- 19:42Sensor Quality
- 25:58P300 ERP
- 30:42Azure ML
- 31:58Azure ML Template
- 41:38Visualizing the Results
- 44:05The Story
- 48:54Data normalization
- 53:00Feature extraction
- 58:22Next steps
- 59:18Thank you
- 59:46Summary
related talks
01h 05mUsing EEG and Azure Machine Learning to Perform Lie Detection
31mJennifer Marsman, Principal Developer Evangelist, Microsoft
26mBrainDrain Using Machine Learning and Brain Waves to Detect Errors in Human
35mTranslating Brain Electrical Signals into Actions through Models...
11mConnected brains
46mKeynote: How Can We Trust Machine Learning?
53mBrain: The ultimate hack
29mAI & Neurobiology of Deception: How AI Helps Us Understand Ourselves