Abstract FitMirror is a smart mirror which recognizes a total of 10 fitness excercises via Mircrosoft Kinect and a Wii Balance Board. The FitMirror supports the users' fitness programm by giving them fitness exercises in the morning and motivating them for the day. By visualizing the users' statistics we help them control their progress.

Concept The mirror supports speech and touch input. Fitness data is transfered to the system via the Google Fit API and our Android App. For also achieving a psychological effect rather than just physical challenge, we offer 5 exercises in the form of games and 5 classic fitness exercises.

In order to compare oneself with friends there is a score calculation of the scored repetitions in relation to the executed time as well as the possibility to challenge friends with the own score. Furthermore an automatic raise of difficulty in form of new targets encourages ones motivation. An own statistic offers an overview of past steps, burned calories and history of weight in order to keep in sight ones proceedings. Since the system is a mirror the UI is limited to a minimum to present the most important things the best way. As a result of this you hold the possibility of observing yourself while doing the exercises in order to recognize your straight back.

Highlights

• Useridentification by smartphone
• 10 different exercises
• 5 exercises in form of games
• Challenge own friends
• Motivation by new calculated targets

Technology To use the touch functions and the spy-foil at the same time, we are using an infrared-touchpanel with a spy-foil on the 42'' monitor. To recognize the body a Kinect is used. In addition the Balance Board with its 4 weight sensors is used to measure weight and balancing of the user. Thereby, we are using Bluetooth 2.0 and USB 3.0 SuperSpeed to communicate with the hardware. The communication between the different software-components enables a client-server-network based on Semaineprojects. Furthermore we are using the GoogleFit API to get the fitness data from the user.

The normal interaction of a User and a system is still very explicit. From pressing buttons on a keyboard to issuing commands through a touch interface.
Recent development by big companies like Google, Apple and Microsoft tries to tackle this problem with more intelligent interaction Methods like voice commands and speech-to-intention recognition.

We present a home system which identifies and tracks the users ubiquitously. The System is build from easy to get, affordable Sensors and Hardware. It is able to resolve ambiguous situations and  able to analyze conflicting situations.
We use the systems as a backbone to create a ubiquitous multimedia service. If a user starts a music playback, the service redirects the multimedia stream to the current room the user is in, allowing for a 'decoupled' multimedia consumption, even managing multiple users at the same time.

The following technologies were used in this project:

mDNS - http://www.multicastdns.org
DNS-SD - http://www.dns-sd.org/
RTP Multicast - http://www.ietf.org/rfc/rfc3550.txt
Node.js - https://nodejs.org/
Deepstream.io - https://deepstream.io/
Nools - http://c2fo.io/nools/
Raspberry Pi - https://www.raspberrypi.org/