This project builds a K-Nearest Neighbours Algorithm to classify phonemes from a data-set.
Given a very large dataset of speech from the British Isles, all were manually inputted into the .csv file attached by generating an average value for each of the 3 formants present in the audio.
The audio was cut up and analysed using the free-ware Praat
The scope of the project was to be able to classify the phonemes from one another. Therefore, a KNN was built in order to do just that. Resulting in an average accuracy of 89%, the notebook with the code can be found in the repository as 'Phoneme Recognition.ipynb'
