Speech aid for the deaf based on a representation of the vocal tract: the vowel module

Presents a new method for speech training of deaf and hard-of-hearing people, using a computer accompanied with an A/D sound card. The idea is to produce a visual feedback of the deaf person´s voice, also giving information about what must be done in order to correctly produce the required speech sound. The visual representation attempts to reproduce the configuration of the speaker´s vocal tract in real time, during vowel articulation. Since changes in this configuration reflect changes of the spectral envelope of the speech signal, we need to solve the inverse problem: obtaining the vocal tract shape from the speech signal. To solve this problem, signal processing techniques such as linear predictive coding were used. We hope that this representation can be more effective than other representations in use today, such as spectra and spectrograms, due to its closer association with the speech production processes