DeepRemote: A smart remote controller for intuitive control through home appliances recognition by deep learning

Abstract

In ordinary households, as the number of home appliances increases, the number of remote controllers also increases. This makes it difficult to manage remote controllers and find an appropriate one for the target appliance. Recently, systems that can control multiple home appliances through one interface have been proposed. However, such systems have the following problems: interface gets complicated as the number of target appliances increases, a special device needs to be attached to each home appliance or a controller. In this paper, we propose DeepRemote, a smart remote controller that can recognize various home appliances and operate them through IR or network. DeepRemote mounts a camera, an infrared (IR) transceiver, Wi-Fi module, four buttons, an inertial measurement unit (IMU) and a speaker. To recognize a target home appliance, the user just points DeepRemote to the appliance intuitively. After recognition, it automatically enters a particular mode to operate the known home appliance. Then the user can control it through a simple interface using buttons and gestures. Through experiments, we confirmed that DeepRemote achieved 81.07% recognition accuracy for five types of home appliances from different user positions, with about 5 seconds for operating a home appliance when the recognition is accurate.