半监督极限学习机用于Android手机活动识别的研究

发布时间：2018-06-05 10:34

本文选题：活动识别 + 半监督极限学习机(SS-ELM)　；参考：《高技术通讯》2017年03期

【摘要】：基于对现有Android手机活动识别技术的分析,针对从不完全、不充分的移动传感器数据中推断人体活动的难题,将能根据无标签样本提高识别预测准确性和速度的半监督(SS)学习和体现模式分类回归的有效学习机制的极限学习机(ELM)相结合给出了解决Android手机平台的人体活动识别问题的半监督极限学习机(SS-ELM)方法,并进一步提出了主成分分析(PCA)和半监督极限学习机(SS-ELM)结合的PCA+SS-ELM新方法。实验结果表明,该方法对人体活动的识别正确率能达到95%,优于最近提出的混合专家半监督模型的正确率,从而验证了该新方法是可行性。
[Abstract]:Based on the analysis of the existing Android mobile phone activity recognition technology, aiming at the problem of inferring human activity from incomplete and insufficient mobile sensor data, Combined with semi-supervised SSSS learning which can improve the accuracy and speed of recognition and prediction based on unlabeled samples, and the extreme learning machine (ELM), which embodies the effective learning mechanism of pattern classification and regression, the human body activity recognition of Android mobile phone platform is presented. The semi-supervised extreme learning machine (SS-ELM) method for other problems, Furthermore, a new PCA SS-ELM method combining principal component analysis (PCA) with semi-supervised extreme learning machine (SS-ELM) is proposed. The experimental results show that the accuracy rate of this method is 95%, which is better than that of the semi-supervised hybrid expert model, which verifies the feasibility of the new method.
【作者单位】：燕山大学信息科学与工程学院;
【基金】：国家自然科学基金(61273019,61473339) 河北自然科学基金(F2013203368) 河北省青年拔尖人才支持计划([2013]17) 河北省博士后专项(B2014010005) 中国博士后科学基金(2014M561202)资助项目
【分类号】：TP181

【相似文献】