The Action Logger
Smartphone multimodal sensor-based activity recognition


Overview
Surveys
Demonstration

Home > Overview
Action Logger Opensource Project

blockDiagram

Motivation
  • To recognize various human activities using multimodal sensors in a Smartphone for providing personalized services.
  • To provide activity recognition core engine for developing smartphone applications utilizing activity information such as fitness, tracking or behavior monitor.
  • To utilize the cutting edge technology for processing sensor raw data from accelerometer and GPS sensor.
Goals
  • Our goal is to recognize 5 different activities which are stay, walking, jogging, taking a bus and subway using accelerometer and GPS sensors in Smartphone.
  • Also provide 3 applications - Fitness Tracking, Human Black Box and Pedometer - based on action logger core engine.
Features
  • Utilizing raw sensor data from accelerometer and GPS sensor in a Smartphone.
  • Recognizing 5 different activities - Standing Still, Walking, Jogging, Talking a bus and Subway.
  • Calculating consumed calorie of each activities using METs index.
  • Providing activity recognition core engine as Java library for developing smartphone application utilizing human's activity information.
  • 3 example applications using action logger core engine are provided.
  • Novel feature extraction algorithms using SVM(Signal Vector Magnitude) and FFT are utilized for processing accelerometer data.
  • Classification algorithm using GMM(Gaussian Mixture Model) is utilized for training and classification phase.
Uniqueness
  • Action Logger utilizes only sensors embedded in a Smartphone - Acceleormeter and GPS.
  • Collecting, Storing and Processing gathered sensor data are handled on Smartphone platform.
  • Calculation of calorie consumption of each activities precisely.
  • The core engine is released as open source for smartphone application developers.