상세정보
학술대회
Activity Recognition using Fully Convolutional Network from Smartphone Accelerometer
- 발행일
- 201810
- 출처
- International Conference on Information and Communication Technology Convergence (ICTC) 2018, pp.1482-1484
- 협약과제
- 18HS1800, 신체기능의 이상이나 저하를 극복하기 위한 휴먼 청각 및 근력 증강 원천 기술 개발, 신형철
- 초록
- This paper presents an activity recognition using smartphone built-in accelerometer. One of the most important issues in implementing activity recognition on embedded systems, including smartphones, is to achieve a high accuracy with a low computational cost and low memory usage. In this paper, we propose an activity recognition using the fully convolutional networks and introduce a new method to generate an input signal image using the combination of deep features and orientation-independent features. The experimental results show that the proposed method is able to achieve a high accuracy with a low memory usage.
- KSP 제안 키워드
- Activity Recognition, Built-in, Embedded system, Fully Convolutional network, High accuracy, Input signal, Low Computational Cost, Low memory usage, Orientation-independent, Smartphone accelerometer, deep features