Activity Recognition with Deep Learning

WISDM v1.1 Activity Prediction Dataset [1] [2]: http://www.cis.fordham.edu/wisdm/dataset.php

Jupyter notebooks can be run directly provided dependencies are installed. All experiment material is included in various subdirectories as detailed below:

data

• cleaned_WISDM_v1

Contains cleaned data as generated by DataUnderstanding.ipynb

logs

    Contains logs from experiments. Both per epoch and evaluation results were recorded for each.

model_training_scripts

• best_cnn3_all_conv_5_fold.py
• best_cnn3_all_conv_hybrid_5_fold.py
• best_cnn3_all_conv_lstm3_5_fold.py
• best_cnn3_all_conv_lstm3_hybrid_5_fold.py

Scripts for each model used in experiments. Scripts perform 5 fold cross validation. For information on motivation for using particular architectures and hyperparameters see [3][4].

src

• utils.py

  Necessary functions for model scripts.
  

• reporting.py

  Code for plotting confusion matrix
  

tests

• test_utils.py

  Unit tests for utils class
  

CNN_LSTM_WISDM.ipynb

Loads and runs final CNN-LSTM models on impersonal data.

CNN_LSTM_WISDM_Hybrid.ipynb

Loads and runs final CNN-LSTM models on hybrid data.

CNN_WISDM.ipynb

Loads and runs final CNN models on impersonal data.

CNN_WISDM_Hybrid.ipynb

Loads and runs final CNN models on hybrid data.

DataUnderstanding.ipynb

Downloads data, shows various attributes of data and displays entire data preparation process. Exports cleaned_data into data directory.

Dependences

Note: Older/newer packages may work but this cannot be gauranteed:

• Python==3.5.4
• pandas==0.20.3
• numpy==1.13.3
• Keras==2.1.2
• tensorflow==1.4.1
• scikit-image==0.13.0
• scikit-learn==0.19.0
• matplotlib==2.0.2.
• jupyter==1.0.0

** For DataUnderstanding.ipynb wget is required to download dataset should it be required.

References

[1] Kwapisz, J.R., Weiss, G.M., Moore, S.A.: Activity recognition using cell phone accelerometers. ACM SigKDD Explorations Newsletter 12(2), 74–82 (2011)

[2] Lockhart, J.W., Weiss, G.M., Xue, J.C., Gallagher, S.T., Grosner, A.B., Pulickal, T.T.: Design considerations for the wisdm smart phone-based sensor mining ar- chitecture. In: Proceedings of the Fifth International Workshop on Knowledge Discovery from Sensor Data. pp. 25–33. ACM (2011)

[3] Ord́oñez, F.J., Roggen, D.: Deep convolutional and lstm recurrent neural networks for multimodal wearable activity recognition. Sensors 16(1), 115 (2016)

[4] Ronao, C.A., Cho, S.B.: Human activity recognition with smartphone sensors using deep learning neural networks. Expert Systems with Applications 59, 235–244 (2016)