Darwin Phones: the Evolution of Sensing and Inference on Mobile Phones

Darwin Phones: the Evolution of Sensing and Inference on

Mobile Phones

Emiliano Miluzzo†, Cory T. Cornelius†, Ashwin Ramaswamy†, Tanzeem Choudhury†,

Zhigang Liu§, Andrew T. Campbell†

†Computer Science, Dartmouth College, Hanover, NH, USA

§Nokia Research Center, 955 Page Mill Road, Palo Alto, CA, USA

ABSTRACT

We present Darwin, an enabling technology for mobile phone

sensing that combines collab orative sensing and classification techniques to reason ab out human b ehavior and context on mobile phones. Darwin advances mobile phone sensing through the deployment of efficient but sophisticated

machine learning techniques sp ecifically designed to run directly on sensor-enabled mobile phones (i.e., smartphones).

Darwin tackles three key sensing and inference challenges

that are barriers to mass-scale adoption of mobile phone

sensing applications: (i) the human-burden of training classifiers, (ii) the ability to p erform reliably in different environments (e.g., indo or, outdo or) and (iii) the ability to scale

to a large numb er of phones without jeopardizing the “phone

exp erience” (e.g., usability and battery lifetime). Darwin is

a collab orative reasoning framework built on three concepts:

classifier/mo del evolution, mo del p o oling, and collab orative

inference. To the b est of our knowledge Darwin is the first

system that applies distributed machine learning techniques

and collab orative inference concepts to mobile phones. We

implement the Darwin system on the Nokia N97 and Apple iPhone. While Darwin represents a general framework

applicable to a wide variety of emerging mobile sensing applications, we implement a sp eaker recognition application

and an augmented reality application to evaluate the b enefits of Darwin. We show exp erimental results from eight

individuals carrying Nokia N97s and demonstrate that Darwin improves the reliability and scalability of the pro of-ofconcept sp eaker recognition application without additional

burden to users.

Categories and Sub ject Descriptors: C.3 [Special-Purpose

and Application-Based Systems] Real-time and embedded

systems

General Terms: Algorithms, Design, Experimentation,

Human Factors, Measurement, Performance

Keywords: Mobile Sensing Systems, Distributed Machine

Learning, Collaborative Inference, Mobile Phones

Permission to make digital or hard copies of all or part of this work for

personal or classroom use is granted without fee provided that copies are

not made or distributed for profit or commercial advantage and that copies

bear this notice and the full citation on the first page. To copy otherwise, to

republish, to post on servers or to redistribute to lists, requires prior specific

permission and/or a fee.

MobiSys’10, June 15–18, 2010, San Francisco, California, USA.

Copyright 2010 ACM 978-1-60558-985-5/10/06 ...$10.00.

1. INTRODUCTION

The continuing need to communicate has always pushed

people to invent better and more efficient ways to convey

messages, propagate ideas, and share personal information

with friends and family. Social-networking, for example, is

the fastest growing phenomenon of the Internet era where

people communicate and share content with friends, family,

and acquaintances. Recently, researchers started investigating new ways to augment existing channels of communication and improve information exchange between individuals using the computational and sensing resources offered

by sensor-enabled mobile phones (aka smartphones). These

phones already utilize sensor data to filter relevant information

...