How to use the code:

Note: The latest version of the code can be found under the branch master.

NLP

To run the preprocessing of the audio data and transcribe each file, follow these steps:

• Install the environment using the VR.yml file.
• In the terminal, run the file audio.sh to executes the overall pipeline.
  • Store the raw data in a folder called data/raw_audio_data If you decide to store the data in a different location, please pass it as an argument to audio.sh. For example, if your data are stored in home/user/data, run in the terminal bash audio.sh home/user/data

Audio preprocessing

The preprocessing of the audio files is done by running the script audio_preprocessing.py The script parameters are the following:

• input_dicPath to the directory containing the raw audio data.
• output_dic Path to directory to save clean audio files and csv files. (These files are created automatically)

Output:

• Processed audio files (along with csv files listed below) are stored in the folder data/clean_audio_data.

• clean_audio.csv contains the participant's id, path to the (clean) audio files and audio's length.

• audio_not_processed.csv stores empty files as well as files that do not follow the naming convention: id_ConditionMode_station. These files must be checked manually.

Transcription

The transcription of the audio files is done by running the script transcription.py The script parameters are the following:

• input_csv Path to csv file containing clean audio files.
• output_csv Path where the csv file containing the transcription is saved.
• output_chunks Path where the long-file's chunks will be stored.
• l or --language Indicate the language of transcription (English or German). Please changeaudio.sh accordingly.

The transcriptions are stored in data/transcription/transcription.csv

Processing transcription

To shuffle the transcriptions and make them anonymous, run transcription_preprocessing.py with the following parameters:

• transcription Path to the csv file containing the transcriptions to be anonymized.
• num_csvs_chunks Number of chunks. E.g., if you have 10 annotators you may want to chunk the transcriptions into 10 portions.
• output_dir Path to the output directory where the resulting data will be saved.

In order prepare the data (transcriptions along with the ratings given by the annotators) for the analysis, run transcription_preprocessing.py with the following parameters:

• path_rated_transcription Path to the csv file that contains the anonymized transcriptions along with their ratings.
• output_dir Path to the output directory where the resulting data will be saved.

Example (For Linux users) python -m NLP.transcription_preprocessing --path_anonymized_transcription /path/to/annotations/csv_file.csv --output_dir /path/to/store/results

Postural stability test

To run all the steps needed to compute the average velocity run post_stability.sh. Pass the path to the data as a parameter. E.g., bash post_stability.sh home/user/data After running the post_stability.sh two files are generated: timestamps.csv and velocities.csv. You can fine them here: data/post_stability. If you want to save the resulting csv files to a different directory, please follow the steps below.

Live data preprocessing.

The preprocessing of the data, which required to compute the average velocity, is done by running the script post_stability_dataprocessing.py The script parameters are the following:

• input_dir Path to the input directory that contains the data.
• output_dir Path to the output directory where timestamps.csv and velocities.csv will be saved