A simple speech recognition system using Linear predictive coding (LPC) and tested on spoken digits
The available data (which you can complete with your own records) consist of audio recordings of people pronouncing numbers (from 0 to 9):
- 4 people
- 2000 registrations in total (50 registrations of each digit per person)
- English pronunciation The file names are in the form of'chiffre_nom_index.wav', for example 1_theo_10.wav'.