Training human activity recognition for labels with inaccurate time stamps
| Autor: | Naonori Ueda, Sozo Inoue, Takamichi Toda, Shota Tanaka |
|---|---|
| Rok vydání: | 2014 |
| Předmět: |
Computer science
business.industry Association (object-oriented programming) Supervised learning Pattern recognition Machine learning computer.software_genre Activity recognition Naive Bayes classifier Action (philosophy) Expectation–maximization algorithm Artificial intelligence business computer Smoothing |
| Zdroj: | UbiComp Adjunct |
| DOI: | 10.1145/2638728.2641297 |
| Popis: | We generally use supervised learning when performing activity recognition using mobile sensor devices such as smartphones. In this application, case data associated with the sensor information and type of action is required. However, there is a possibility that a time shift occurs because this association is done manually on the audio and video that has been acquired along with the sensor information. In this paper, we propose a method of activity recognition that can recognize correct actions even if there is a time gap. In this method, we add labels that shift the original learning data label. We also implement multi-label machine learning. In addition, we propose a method for repeated learning based on the Expectation-Maximization(EM) algorithm. To evaluate this method, we conducted an experiment that recognized three types of behavior using a Naive Bayes classifier. In the evaluation, we pieced together three types of human action data into one dataset called pseudo sequence data. We slid the action labels of the pseudo sequence data and examined whether the recognition rate was improved by our proposed method. The results show that the proposed method can perform activity recognition with high accuracy, even if the action labels times are shifted. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|