A computing device is described that uses motion data from motion sensors to detect user inputs, such as out-of-screen user inputs for mobile devices. In other words, the computing device detects user touch inputs at locations of the device that do not include a touch screen, such as anywhere on the surface of the housing or case of the device. The techniques described enable a computing device to utilize a standard, existing motion sensor (e.g., an inertial measurement unit, (IMU), accelerometer, gyroscope, etc.) to detect the user input and determine attributes of the user input. Motion data generated by the motion sensor (also referred to as a movement sensor) is processed by an artificial neural network to infer characteristics or attributes of the user input, including: a location on the housing where the input was detected, a surface of the housing where the input was detected (e.g., front, back, and edges, such as top, bottom and sides); a type of user input (e.g., finger, stylus, fingernail, finger pad, etc.). In other words, the computing device applies a machine-learned model to the sensor data to classify or label the various attributes, characteristics, or qualities of the input. In this way, the computing device utilizes machine learning and motion data to classify attributes of the user input or gesture utilizing motion sensors without the need for additional hardware, such as touch-sensitive devices and sensors.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Quinn, Philip; Huang, Michael Xuelin; and Zhai, Shumin, "DETECTING ATTRIBUTES OF USER INPUTS UTILIZING MOTION SENSOR DATA AND MACHINE LEARNING", Technical Disclosure Commons, (July 10, 2019)