A brain computer interface system interprets an individual's neural signals to predict specific phonemes, words, or sentences, thereby enabling the individual to communicate with others through unspoken methods. Specifically, a brain computer interface system captures neural signals from an individual at mesoscopic resolutions using optical neuroimaging techniques. The system applies the captured neural signals to multiple predictive models that have been trained on neural signals captured from previous individuals. The predictive models output predictions as to a phoneme or word that corresponds to the captured neural signals. Therefore, the individual can communicate through the brain computer interface system by only providing neural signals without using verbal, expressive, or physical means.