VibroTactor: Low-cost Placement-Aware Technique using Vibration Echoes on Mobile Devices Sungjae Hwang, Kwang-yun Wohn Graduate School of Culture Technology Korea Advanced Institute of Science and Technology {best, wohn}@kaist.ac.kr ABSTRACT In this paper, we present a low-cost placement-aware technique, called VibroTactor, which allows mobile devices to determine where they are placed (e.g., in a pocket, on a phone holder, on the bed, or on the desk). This is achieved by filtering and analyzing the acoustic signal generated when the mobile device vibrates. The advantage of this technique is that it is inexpensive and easy to deploy because it uses a microphone, which already embedded in standard mobile devices. To verify this idea, we implemented a prototype and conducted a preliminary test. The results show that this system achieves an average success rate of 91% in 12 different real-world placement sets. Author Keywords pseudo sensor; sensor repurposing; pattern recognition; placement detection; context-aware; vibration echoes. ACM Classification Keywords H.5.2 [Information interfaces and presentation]: User Interfaces - input devices and strategies. INTRODUCTION Context-awareness is an important issue in the research on mobile devices because it can provide richer user experiences by freeing users from the manual instructions of everyday computers. Therefore, numerous context-aware applications have been introduced. One common real-world application is the context-adaptive graphical user interface, where, for instance, the screen layout and brightness of mobile devices change based on the environmental conditions (e.g., orientation and ambient light). Another example is the LBS (location-based service). This service is now widely used outdoors, relying on GPS, as well as indoors through its use of a signal-strength sensing technique (e.g., wireless access points, cellular signals, Bluetooth beacons [6], RFID [5], and geo-magnetism [3]) However, these techniques require the environment to have the proper infrastructure, and they only give information about a small part of various user contexts (coarse localization). To address this issue, researchers have introduced localization techniques that utilize specialized sets of sensing hardware. Harrison et al. [1] introduced a placement-aware sensing technique for mobile devices that showed a placement-detection accuracy rate of 86.9% for 27 different test placements. To infer the placement, they used five additional material sensors (an ultraviolet LED, a RGB LED, a photo resistor, a light sensor IC, and an infrared LED). This technique offers reliable and fast (or real-time) localization, but one disadvantage is that such hardware is still not supported in commercially available mobile devices. Given this background, some researchers have investigated the inertial sensors of mobile devices, such as their microphones or accelerometers [2, 4]. Cho et al. [2] introduced a surface recognition technique that used a built-in vibration motor and accelerometer. They used the vibration signature in conjunction with the inertial accelerometer in their device. Their system achieved accuracy rates that exceed 85% in six places (a sofa, a plastic table, a wooden table, the user’s hands, in a backpack, and in the pocket of a pair of pants) Kunze et al. [4] proposed a symbolic localization method that worked by the active sampling of vibration and sound signatures. They sampled vibration and short frequency ‘beeps’ to discover symbolic locations through inertial sensors such as an accelerometer and microphone. Although they achieved recognition rates of up to 78% for 35 semantic locations by combining several methods, a rate of only 51% was achieved when using vibration sound. In this paper, unlike previous approaches, we considered and focused on using a single built-in sensor, a microphone, to recognize the types of surfaces on which the device is placed. PROPOSED METHOD To infer the placement of the device, we exploited the simple phenomenon in which different locations generate different vibration echoes. For instance, the vibration of a mobile device on hard materials (e.g., wood, glass, and stone) emits noticeable sounds to the user. However, vibrations on soft materials (e.g. fabric, rubber, and sponge) do not emit noticeable sounds. Figure 1 shows different spectrograms for 12 different placements. The difference between the spectrograms of the desk and the bed is distinguishable to the naked eye. To verify our concept, we implemented a prototype on Android 2.3.6 (on a Samsung Galaxy S2). We used JNI with the FMOD library to perform a fast Fourier transform (FFT) on incoming signals Copyright is held by the author/owner(s). IUI’13 Companion, March 19–22, 2013, Santa Monica, CA, USA. ACM 978-1-4503-1966-9/13/03.