HAR滑动窗口预处理(滑窗切割)及网络搭建(训练)实战

Human Activity RecognitionHAR)译文:人类活动识别,人类行为识别,人体姿态识别

时序数据的预处理方法包含很多,其中滑动窗口预处理为增加数据量的关键,这里自行整理了7个公开数据集的滑窗预处理方法,形成可送入CNNResNet网络的数据形式用于HAR任务。

至于其他的预处理方法,例如标准化,缺值填充等可以根据需求自行在代码中添加

Daily-and-Sports-Activities-dataset:HAR-Dataset-Preprocess/Daily_and_Sports_Activities_dataset at main · xushige/HAR-Dataset-Preprocess · GitHub

OPPORTUNITY Dataset:HAR-Dataset-Preprocess/OPPORTUNITY at main · xushige/HAR-Dataset-Preprocess · GitHub

PAMAP2 Dataset:HAR-Dataset-Preprocess/PAMAP2 at main · xushige/HAR-Dataset-Preprocess · GitHub

UCI-HAR Dataset:https://github.com/xushige/HAR-Dataset-Preprocess/tree/main/UCI_HAR

The USC-SIPI Human Activity Dataset(USC-HAD):https://github.com/xushige/HAR-Dataset-Preprocess/tree/main/USC_HAD

UniMib-SHAR Dataset:https://github.com/xushige/HAR-Dataset-Preprocess/tree/main/UniMib_SHAR

WISDM Dataset:https://github.com/xushige/HAR-Dataset-Preprocess/tree/main/WISDM

你可能感兴趣的:(python,神经网络,深度学习,cnn,分类)