中文摘要：

提出了一种基于多传感器可穿戴设备的无监督室内/室外场景的区分方法。首先,该方法对多维的传感数据进行时间序列建模,通过分析该时间序列挖掘出场景切换的模式并对该时序数列进行分段分析;接着,建立相似性测量模型对每个分段时间序列进行室内/室外场景相似度计算,根据计算的结果识别出室内/室外场景。通过实验分析,该方法对室内/室外场景区分准确度高达90.1%,相较于其他方法准确度提高了13%~33%。该方法无须对数据人工标记,实现了较高的场景区分准确率,适用于大规模数据采集场景。

英文摘要：

This paper presented an unsupervised method for indoor/outdoor （IO） detection in environmental sensing via wea- rable devices equipped with multiple sensors. Firstly, the method constructed the context model on multi-dimensional time se- ries sensor data. And it captured the context switching patterns （CSP） from the perspective of local trends resulted from IO context switching. Secondly,it built a simple yet efficient model to measure the similarity between a multi-dimensional outdoor reference time series and the sensor data collected by the wearable devices. As a result, the data could be classified into in- door and outdoor categories by the proposed model. It validated this method in a real environment with the sensors. The result shows that the classification accuracy of the proposed method is about 90.1% which is 13% ~ 33% better than other alterna- tives, The method doesn＇ t require manually labeled training datasets. It achieves classification with high precision during the large scale collection.