Synthetic Video Model for Hearing-Impaired Speech Correction

The patent introduces visual feedback as an intermediary to bridge the gap caused by hearing impairment. A camera captures the user's mouth movements and compares them to correct articulation models, providing visual feedback that compensates for the lack of auditory feedback. This intermediary mechanism allows hearing-impaired individuals to learn speech correction without relying on hearing.

The patent replaces the auditory feedback mechanism with a visual feedback system. Instead of using sound waves to convey pronunciation information, the system uses visual images of mouth articulations captured by a camera and processed by image recognition algorithms. This substitution transforms the feedback modality from acoustic to optical, making it accessible to hearing-impaired users.

The patent creates visual copies of correct mouth articulations and compares them with the user's actual articulations. The system captures images of correct pronunciations, processes them through image recognition to extract feature points, and compares these features with the user's speech images. This copying and comparison approach enables precise measurement of articulation accuracy without requiring overly complex analysis systems.

The patent divides the speech analysis task into discrete segments - capturing individual phonemes or syllables rather than analyzing continuous speech. The image recognition system processes speech images by identifying key feature points (lips, cheeks, jaw) in segmented frames, making the complexity manageable by breaking down the continuous speech stream into analyzable discrete units.