A smartphone user's biometric information, which can represent the individual, is useful for customized mobile marketing and advertising, and also be used in health-care applications for personalized services. In this study, we propose a method based on clustering and deep learning to infer the height and weight of a smartphone user by using multimodal sensor data in various moving situations. Our method handles data collected under relaxed conditions compared to previous work, in that the

more »

... ion of a smartphone, and the user's ways and means of transportation are not limited. A clustering-based data selection method is presented to filter data helpful for inferring biometric information. Furthermore, a recurrent neural network, which has been successful in human-activity 군집화와 딥러닝을 활용한 스마트폰 센서 데이터로부터의 신장과 체중 추론 기법 457 recognition, is adopted to improve the performance. We collected smartphone sensor data from 26 users in their actual living environment with a self-developed application, and our experimental results confirmed the effectiveness of the proposed method.