Used Multinomial Logistic Regression (Softmax Regression) for Gender Classification and Speaker Recognition using speech.
In phase I of the project, multiple recordings were obtained from each student in the class. In phase II, the Multinomial Logistic Regression was implemented from scratch. In phase III, scikit-learn’s ‘Multi-layer Perceptron’, ‘Linear Support Vector’ and ‘Gaussian Naïve Bayes’ implementations were used to train and test the multi-layer neural networks, linear SVM, and Gaussian Naive Bayes, respectively, on the data.
Trained, validated, tested the model and reported evaluation metrics on the test set.