This is interesting. My initial thought is that it would try to uniquely identify the sound of each individual key, then it would apply statistical analysis to

This is interesting. My initial thought is that it would try to uniquely identify the sound of each individual key, then it would apply statistical analysis to map each key to a letter (the letter 'e' should be most common in English).