Skip to content

junxiant/Detect-Alzheimers-using-Speech-and-Text

Repository files navigation

Alzheimers Detection

  • Total of 4 notebooks, each for different parts of the project
  • Notebooks has different sections to organize codes

Notebooks

  • Generate_Transcripts.ipynb

    • Contains the code used to generate text transcripts using HuBERT
  • Audio_Classification.ipynb

    • Install dependencies: Install openSMILE

    • File preprocessing functions Change labels to integer

    • Functions to extract features: Functions extract eGeMAPS features

    • Extracting features: Extracting eGeMAPS features and saving to a csv file

    • Load features and data preprocessing: Loading features from csv file and performing normalization

    • Machine Learning classifiers: This section has functions to train different classifiers with 10 fold grid search cv (Decision Tree, Decision Tree Bagger, Random Forest, Random Forest Bagger, Support Vector, Logistic Regression)

    • Neural Network: This section contains codes used to create the 4-layer network and do 10 fold grid search cv

  • Text_Classification.ipynb

    • Load text transcripts: Load the generated text transcripts

    • Data Preprocessing: Text data preprocessing, to lowercase all text

    • Doc2Vec: Contains codes used to perform grid search 10 fold cross validation for machine learning classifiers and a neural network using Doc2Vec embeddings

    • BERT: Contains codes to load BERT model, extract embeddings, training and tuning of neural network model

  • Fusion_Mechanisms.ipynb

    • Load pre-trained text model: Loading the text data and some preprocessing, and loading the text model

    • Load pre-trained audio model: Loading audio data and audio model

    • Get audio and text inputs: More data preprocessing and to get audio and text data inputs

    • Bilinear Pooling: Has classes to create a bilinear pooling layer, function to create the bilinear pooling model, and training of the model with 10 fold cv

    • Concatenation: Has function to create the model with concatenate layer and training of the model with 10 fold cv

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published