由于惯性传感器所采集到的原始数据是被采集者运动一段时间的连续数据而采集的原始数据既有动作数据也有非动作的数据。如果将所有的原始数据进行特征提取来进行后续的分类那么非动作的数据将影响特征的提取和分类因此进行特征提取之前要将运动的动作窗口数据截取出来。所谓的窗口分割就是将采集的较长的原始运动数据通过一定的规则分割成具有相同长度数据或不同长度的数据窗口。窗口数据截取的标准是在截取的窗口中只能包含一个完整

Inertial sensors collect continuous raw data over a period of time, which includes both action and non-action data. Extracting features from all the raw data for subsequent classification will be affected by the non-action data, hence the action window data needs to be extracted before feature extraction. Window segmentation refers to dividing the long raw motion data into data windows of the same or different lengths according to certain rules. The standard for window data extraction is that each window should only contain complete action data and exclude data unrelated to the action as much as possible