This repository contains a series scripts for extracting features from .wav files - these scripts were written to better understand Natural Language Processing.
Program that makes use of the nltk (Natural Language ToolKit) library to parse through a text file and construct unigrams and bigrams based off of word frequencies. Additionally, by examining the rate at which certain words appear within a text, the program can use a greedy algorithm to predict the most probable sentences to appear within the text.
Program that utilizes the concepts of Dynamic Time Warping by parsing through two text files of sound features, where each line of a text file will represent one frame of a .wav file (with each value representing a particular feature). By comparing two files against each other, one can make an assumption about whether or not the two files are representing the same word or phrase; the smaller the Euclidean distance, the more likely it is that the two text files are representing the same word/phrase.
Program that relies on several libraries include sklearn (SciKit-Learn) and NumPy to generate classifiers that can be used to evaluate the speech features of a speaker. By examining an individual's speech features, assumptions can be made about whether a speaker has a particular accent or even a speech impediment (such as one that is caused by dementia).
Jordan Edward Shea, Emily Prud'hommeaux