Jennifer Marsman, Principal Developer Evangelist, Microsoft
summary
Uses an Emotiv EPOC headset to capture 14‑channel EEG at 128 Hz sampling, maps signals to actions (a virtual cube moves reliably via mental command) and labels time slices as truth/lie to train classifiers in Azure Machine Learning. Two algorithms tested — logistic regression and a decision jungle — with the jungle showing much better accuracy on a single‑subject dataset (husband). Notes noisy facial artifacts, sensor‑quality filtering, marker timing limits, and ongoing work to generalize across people.
AI-written summary · about
chapters (16)
- 0:00EEG and lie detection intro
- 3:37EEG and Emotiv headset
- 6:26EEG applications beyond lies
- 9:46Mind-controlled cube demo
- 10:51Lie detection experiment design
- 13:22Annotating EEG data with Test Bench
- 14:44EEG data cleaning and sensors
- 16:51Existing EEG lie detection research
- 17:11Azure ML overview
- 20:46Choosing ML algorithms in Azure
- 23:13Preparing EEG data for Azure ML
- 24:36Training lie detection models
- 26:20Evaluating model results
- 27:28EEG lie detection challenges
- 29:58Next steps and future experiments
- 31:04Closing remarks
related talks
01h 05mUsing EEG and Azure Machine Learning to Perform Lie Detection
01h 01mKeynote: Using EEG and Machine Learning to Perform Lie Detection
46mControlling apps with your mind
35mTranslating Brain Electrical Signals into Actions through Models...
11mConnected brains
24mControlling Apps with Your Mind and AI
53mBrain: The ultimate hack
19mNeuro Game Experience: Your Thought Takes Control!