[English / Japanese]

LetterTwitter: Smart Mailbox for Spam-filtered Notification of Received Letters
Last Updated: 2010/10/09


Abstract

We propose a smart mailbox called "LetterTwitter" that can automatically capture dropped snail mail (s-mail), classify pictures into several categories (e.g., letters or flyers), and upload categorized pictures to the Web. Users can easily get spam-filtered notification of received letters using PCs or cellular phones equipped with common web browsers.
Figure 1. The concept of LetterTwitter:
Spam-filtered notification of received letters appears on the Web. (demo video)


Imprementation

Before installing the system to an actual house, Ocha House, we developed a prototype of the same size (40 cm x 40 cm x 30 cm) as the mailbox in the house (Fig. 2, left). The prototype mailbox was assembled from acrylic boards. We attached two magnetic switches to the front and back slot covers, two USB cameras (Logicool Qcam Pro for Notebooks) to the front/outside and top/inside of the mailbox, and a small fluorescent lamp to the top/inside (Fig. 2, right). We also attached a small PC inside the mailbox.

The system aims to classify s-mail into several categories. In the current prototype, we have implemented the function to distinguish flyers from letters. We focus on the difference in the color distribution between letters and flyers; that is, while most letters have low saturation (e.g., white), most flyers tend to have high saturation (e.g., primary colors) (Fig. 3). Therefore, we decided to classify s-mail by the ratio of high saturation pixels in each picture.Please read our paper for details.

Figure 2. The prototype of LetterTwitter
(left: front-side, right: device architecture).

 Figure 3. Examples of captured pictures. While letters often have low saturation (top), flyers tend to have high saturation (bottom).


Installation
After the prototyping, we have installed the system in Ocha House, as shown in Fig. 4. We are currently collecting data on the accuracy of classification, intervals of s-mail arrival, and feedback from users for further improvements.
Figure 4. LetterTwitter installed in the Ocha House


Demo video
Full Version (2min 25sec)
High resolution version (QuickTime format)


Publication
  • Koji Tsukada, Yuka Mizushima, Ai Ogata and Itiro Siio: LetterTwitter: Smart Mailbox for Spam-filtered Notification of Received Letters, Adjunct Proceedings of Ubicomp 2010 (Video/Poster), pp. 439-440 (Sept, 2010).[PDF]

Koji Tsukada