LSTM_Zero_Velocity_classification

This is project to classify zero-velocity status using the data from accelerometer and gyroscope.

To get processed data

Link: https://mega.nz/folder/89QDxKbR Password: Oy-xsUB8FyFqTjhzy8qYDg

To start

We suggest you run

pip install -r requirements.txt

config.py contains the parameters used in training.You should write your own config.py and put it in the root directory.

Here is an example of config.py:

LR = 0.0003
DATA_DIR = "./data_process/int_1_len_24_91_up"
BATCH_SIZE = 600
NUM_EPOCHS = 300
HIDDEN_SIZE = 12
NUM_LAYERS = 2

You may write your own config.py to try different hyperparameters or use different data.

Scripts

• process_data_path.py: Generates the data used in LSTM training.

• train_lstm.py: Train the LSTM network.

• model_test.py: Test the trained model which is saved as "name_you_decide.pkl"

• data_alignment.py: To align the data (Our data has lots of problems, we suggest you directly use the data downloaded from Mega, if you need the original data, please contact us by email: ljyuan@bupt.edu.cn).