Top > Research > Human Behavior Understanding

research

Digital Human for Human Behavior Understanding

researchimg researchimg

Everyday Life Computing

Introduction

Although scientists have worked on sophisticated models of quantum mechanics and cosmology, there is little understanding of the properties and dynamics of everyday life, let alone its standard model. Modeling everyday life requires quantitatively observing it and finding structures in a large amount of observed data.

The recent development of ubiquitous sensing technology enables the in-vivo observation of a total living space, and statistical modeling technology enables the construction of a model from the observed data. By using these technologies we can open the field of science of everyday life.

Sensing Technology

Research scientists at the Digital Human Research Center (DHRC) have developed and commercialized ubiquitous sensors that are usable in an everyday life space, including acoustic sensors that can collect spatio-temporal data of behavior at the precision of sub-mm to cm and wearable sensors for electromyography and acceleration data.

A sensor home has been built that consists of a bedroom, a living room, and a kitchen, saturated with these sensors. The home has collected over 100-hours of behavioral data of children who play and interact with their parent in it.

These observations of human behavior in the sensor home provide data of everyday life from the microscopic perspective. The Internet works as a sensor, providing data from the macroscopic perspective. An example of such Internet sensing is an injury surveillance system that the DHRC installed at the two children's hospitals in Japan. Since 2005, the system has collected data of over 7,000 children injuries at home, including their kind, cause, and cost of treatment.

Modeling Technology

A model of human everyday life must be reusable; it not only represents the statistical characteristics of human's behaviors, but also can predict behaviors given the environment, situations, and conditions. For that purpose, the underlying semantic causality among them must be adequately modeled.

An everyday-life model of children was developed from the data collected in the sensor home by using BAYONET, a Bayesian-network software tool to support probabilistic modeling in order to express the causal networks of multiple and indeterminate factors. A simulator of children's behavior with graphics display was also developed.

Integration with Service

Integrating sensing and modeling with service allows for sustainable development of the science of everyday life. While providing service we can collect an even larger amount of everyday life data and refine the model, which in turn improves service.

This service integration methodology was applied to many practical information service systems, including personal mobile phones, car-navigation systems, and e-commerce. Its impact is significant. A Web service for supporting precognition of children's home injury that we started in December, 2005, already distributed 44,091 injury animations and collected 14,944 scientifically significant data from 6,052 parents.

researchimg

News

[2004/4 Ultrasonic 3D Tagging System was commercialized]

Research

Everyday Behavior Sensing Technology

Ultrasonic 3D Tagging System

Ultrasonic 3D Tagging System

Technology for Observing Human Activity in the Order of Centimeters

Ultrasonic Radar

Ultrasonic Radar

Minimally Privacy Violative Sensor

Everyday Behavior Modeling Technology

Evidence-based Service Technology

Sensor Network for Elderly Nursing Support

Sensor Network for Elderly Nursing Support

Top ∆