Research Initiation - Optimization of the Visual Vocoder for Speech Therapy

摘要

A TV-display for a visual vocoder has been developed which uses any unmodified TV receiver. It produces a display which is easier to view and capable of more types of displays and storage than the previously used light bulb matrix display. The new display consists of two complete subdisplays, each containing 12 rows of speech frequency and amplitude information, and 32 columns of time information. Four discrete levels of amplitude information are displayed in each row by variation of line width. The columns, which can be split in half to provide a total of four independent displays, have a variation of 16.7-1.33 milliseconds per column. As a result, the final detail of a part of a word or the overall word can be evaluated. Also, the display is capable of storage in any or all of the four possible subdisplays so that a student can attempt to imitate the sound of a teacher. The display is considered an advance over light bulb matrix displays because of improved results of communality tests to measure (1) the sameness of the display when the phoneme is spoken by different speakers, and (2) the uniqueness of the display for different phonemes spoken by the same speaker.