Graphical Phone Representation

The problem of phone recognition is transferred to the visual domain. Each phone is represented by unique graphical Character. The work is applied on a single speaker to eliminate the effect of multi speaker in this phase. The system was trained using handverified data from single speaker. Using 31 context-dependent phone models, a baseline phone accuracy of 63.5% (no phone grammar) has been obtained on an independent test set segments from the same speaker. The evaluation is based on human eye recognition of the generated visual characters.