스마트폰 센서를 이용한 사용자 행위 기반 개인 분류 기법의 성능 평가
Performance Evaluation of Person Classification Techniques Based on Human Activities Using Smartphones' Sensors
0
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
스마트폰의 사용자를 알아내기 위하여 지문, 홍채 등을 사용하는 방법과 같이 별도의 단계를 사용하지 않고, 자연스러운 일상생활의 다양한 행동을 분석하여 사용자를 분류하는 연구가 활발히 이루어지고 있다. 그러나 사람 행동의 다양성과 피실험자 각각에 대한 분석 및 기존 연구의 기법들을 사용한 실험 및 분석은 미흡한 실정이다. 본 논문에서는 스마트폰의 가속도와 자이로스코프 센서를 이용하여 30명의 피실험자로부터 여섯 가지 행동 데이터를 수집하여 공개한 데이터세트를 가지고, 기존 연구 결과에서 우수한 성능을 보인 분류 기법을 가지고 실험하여, 다양한 사람 행동을 가지고 사용자를 분류할 수 있는 방법을 제안한다. 일곱 가지 분류 기법으로 여섯 가지 행동을 비교 실험한 결과, AdaBoost가 우수한 결과를 보여줬다. 다음으로 모든 행동에 대한 분류 실험에서 우수한 결과를 보인 상위 4가지 분류기법을 이용하여 각 행동별 사용자 분류 실험을 한 결과, 눕기, 앉기 및 서기에서는 AdaBoost가 가장 우수하였으며, 걷기, 계단내려가기 및 계단오르기에서는 SVM이 가장 좋은 정확도를 보였고, 앞선 세 가지 동작보다 높은 분류 정확도를 구할 수 있었다. 특히 계단오르기를 이용한 사용자 분류 정확도는 가장 높은 결과를 보였다. 이 실험 결과는 앞으로 지속적 사용자 분류 기술 개발에 중요한 기초가 될 것이라고 생각한다.
Abstract
To distinguish a smartphone user, some studies have been researched to analyze various human activities of natural daily life and classify users without using an additional step, such as using fingerprints or iris. However, it is not sufficient to analyze the diversity of human activities, the analysis per subject and the experiments using techniques that were better in previous related studies. In this paper, we propose a method that classifies users with various human activities by experimenting with a public data set of six activities data collected from thirty subjects using acceleration and gyroscope sensor in smartphone and with seven classification techniques that are better to existing research results. As a result of the comparison experiment with seven classification techniques, AdaBoost showed good results in the experiment in which the user classifies all of the six activities. Next, Adaboost showed the good results in laying, sitting and standing according to each activity by using the top four techniques which were better performance in classification of all activities. In walking, walking downstairs and walking upstairs, we could find out higher accuracy than the previous three activities, and SVM showed relatively excellent accuracy. Especially, the accuracy of user classification using walking upstairs is the highest. The results of this experiment are expected to become an important basis for the development of technologies that naturally and continuously classify users in everyday life.
참고문헌 (16)
- J-H. Kim, "Personal authentication technology and database construction using biosignals", Journal of Telecommunication Technology Association, 165, pp.41~46 (2016)
- J. R. Kwapisz, "Cell phone-based biometric identification", IEEE International Conference on Biometrics: Theory Applications and Systems, pp.1~7 (2010)
- W. H. Lee, "Implicit sensor-based authentication of smartphone users with smartwatch", Proceedings of the Hardware and Architectural Support for Security and Privacy, pp.1~8 (2016)
- M. Ehatisham-ul-Haq, "Identifying smartphone users based on their activity patterns via mobile sensing", Procedia Computer Science, 113, pp.202~209 (2017)
- S. R. Ramamurthy, "Recent trends in machine learning for human activity recognition—A survey", pp.e1254~ (2018), Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery
- M. Janidarmian, "A comprehensive analysis on wearable acceleration sensors in human activity recognition", Sensors, 17(3) (2017)
- D. Anguita, "A public domain dataset for human activity recognition using smartphones", European Symposium on Artificial Neural Networks, pp.437~442 (2013)
- T. B. Singha, "Person recognition using smartphones' accelerometer data", arXiv preprint arXiv:1711.04689, ~ (2017)
- C. Tang, "An empirical evaluation of activities and classifiers for user identification on smartphones", IEEE 8th International Conference on Biometrics Theory, Applications and Systems (2016)
- P. Casale, "Personalization and user verification in wearable systems using biometric walking patterns", Personal and Ubiquitous Computing, 16(5), pp.563~580 (2012)
- T. Hoang, "On the instability of sensor orientation in gait verification on mobile phone", IEEE 12th International Joint Conference on e-Business and Telecommunications, 4, pp.148~159 (2015)
- A. Ignatov, "Real-time human activity recognition from accelerometer data using Convolutional Neural Networks", Applied Soft Computing, 62, pp.915~922 (2018)
- E. Frank, "The WEKA workbench. Online appendix for data mining: Practical machine learning tools and techniques" (2016), Morgan Kaufmann
- R. Kohavi, "A study of cross-validation and bootstrap for accuracy estimation and model selection", International Joint Conferences on Artificial Intelligence, 14(2), pp.1137~1145 (1995)
- G. Shmueli, "Data mining for business intelligence: Concepts, techniques, and applications in Microsoft Office Excel with XLMiner" (2011), John Wiley and Sons
- T. M. Oshiro, "How many trees in a random forest?", International Workshop on Machine Learning and Data Mining in Pattern Recognition, pp.154~168 (2012)