Earthquake Early Warning: MyShake - Building a Global Seismic Network Using Your Smartphones
summary
Project turns smartphones into seismometers and runs a tiny classifier on-device: a three features set (zero crossings, short-window amplitude, cumulative absolute velocity) feeds a 5-neuron network that flags motion; phones send time, location and amplitude and a server clusters triggers requiring 60% within 10 km to declare an event. Team trains on millions of human-activity windows, synthesizes/shake-table earthquakes, logs ~800 offline detections and achieves four real-time alerts so far, with sensor variability and class imbalance still limiting small-quake detection.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 2:25MyShake Technology
- 3:20Challenges
- 3:49Collecting Data
- 4:06Prototype System
- 8:28Shake Table Test
- 10:59Data Overview
- 14:15Workflow
- 16:14Artificial Neural Network
- 17:42Feedback Loop
- 18:51Neural Network
- 20:32Earthquake Signals
- 21:28Feature Selection
- 23:06Feature Selection Schema
- 27:42AI Neural Network
- 34:46Training Data
- 41:01Performance Evaluation
- 48:16Physics
- 51:23RealTime Detection
- 52:43Wrong Decisions
- 53:30Neural Networks
- 54:28Model
- 55:12Questions
related talks
22mBuilding a movement data and analytics platform to de-stress customers & save lives
16mIoT on Your Smartphone: Visualizing Mobile Sensor Data with Grafana
19mLive Build Of A Python Service That Tracks, Transforms & Delivers Sensor Data In Real Time
29mBuilding a mobile location aware system with beacons
17mAI workflow for large scale deployment of far-edge ML devices
21mIoT, Wavelets and Machine Learning in the Smart Home
19mDeus EM Machina: On-Touch Contextual Functionality for Smart IoT Appliances
33mEnterprise Radar