Human Activity Recognition Based on Motion Sensor Using U-Net

---

Y. Zhang, Z. Zhang, Y. Zhang, J. Bao, Y. Zhang and H. Deng, "Human Activity Recognition Based on Motion Sensor Using U-Net," in IEEE Access, vol. 7, pp. 75213-75226, 2019, doi: 10.1109/ACCESS.2019.2920969.

https://ieeexplore.ieee.org/document/8731875

• data
  • OPPGestureDataset
    • features_extraction_codes:provides the codes to conduct the features extractions for machine learing algorithms
    • extracted_features_dataset.txt:provides Baidu Netdisk link to download the extracted features dataset
  • SanitationDataset
    • readme.docx: introudce the Sanitation dataset
    • sanitation.csv:provides the raw dataset
    • Sanitation_processed_data.zip:provides the extracted features dataset
• HAR_dense_prediction_methods:includes the HAR codes of U-Net, FCN, SegNet, MaskRCNN
  • main.py: the main funciton
  • run.py: you can run it directly,configure the parameters
  • postcorrection.py:the source codes of the post correction algorithm
  • unet_info.py: setting the GPU
  • unet_data_load.py: the source codes of the generation of subsequences on the four datasets
  • unet_model.py: the source codes of HAR based on unet
  • FCN_model.py: the source codes of HAR based on FCN
  • segnet.py&layer.py: the source codes of HAR based on SegNet
  • maskrnn.py: the source codes of HAR based on Mask R-CNN
• HAR_sliding_window_prediction_methods: includes HAR based on machine learning and deep learning methods(CNN,lstm,cnnlstm)
  • com_main.py: the main function of HAR based on CNN,LSTM,CovLSTM
  • com_run.py: you can directly run it
  • win_data_load.py:the source codes of the generation of sling window data on the four datasets
  • deep_model_cnn_lstm_covlstm.py: the source codes of HAR based on CNN,LSTM,CovLSTM
  • SVM_nitin.py&decisionTress.py：the source codes fo HAR baesd on SVM and DT
  • common.py: calculate the confusion matrix and Fw-score
  • sameindex.py: convert the sliding window prediction index to the dense prediction index