This Phase I SBIR is prompted by the need for more effective Augmentative and Alternative Communication (AAC) devices for persons unable to communicate through vocalization. The project follows our preliminary work that translates surface electromyographic (sEMG) signals recorded from speech articulation muscles during silently mouthed ?subvocal? speech to produce hands-free intelligible textual communication on an Android or other AAC platform. Now we undertake a first-of-its-kind AAC system that restores personalized, natural expressive vocal communication through an advanced sEMG-based neural interface. This innovation will have broad impact among millions of AAC users by restoring not only the vocabulary, but the prosodic attributes of speech directly to personalized voice synthesis technologies. To achieve this unprecedented goal, our team of experts in sEMG-based subvocal speech recognition at Altec, Inc will partner with VocalID, the world?s leading provider of personalized text-to-voice synthesis for AAC devices. In Phase I we will develop algorithms for tracking sEMG-based changes in prosody at the phrase-level of subvocal speech (Aim 1), synthesize the prosodic changes into personalized digital voice (Aim 2), and establish the proof-of-principle that the sEMG-based digitized voice can communicate phrases with declarative, interrogative and contrastive prosody effectively in people with laryngectomy ? a high-need patient population that lacks the ability to vocalize (Aim 3). Phase II will advance the recognition of intonation, loudness and timing at the triphone-level of subvocal speech and directly synthesize, in real-time, a personalized, intelligible and expressive voice that is capable of communicating meaning, emotion and intent. Our final deliverable will be an AAC system that is unique in restoring acoustic speech for those who live without their natural voice. The device will be packaged as a wearable system that provides natural embodiment, cosmetic appeal, and intuitive vocal capabilities. It will be hands-free, will not suffer from poor intelligibility, will not need surgical intervention or invasive maintenance, and will be readily adaptable to man-machine interface for improved AAC control. The impact of this innovation is that it will provide singular technology that restores personalized and prosodic vocal capabilities to the natural intuitive mechanisms of speech production for the vocally-impaired.
We propose a speech recognition system that uses noninvasive electromyographic (EMG) signals from speech articulation muscles during subvocal (i.e. mouthed) speech to restore the ability to communicate for those with vocal impairments, such as individual?s post-laryngectomy. The EMG Voice Restoration system will improve upon current methods of augmentative and alternative communication (AAC) technologies by offering a non- invasive, hands-free device that synthesizes a personalized, intelligible and expressive voice based entirely on the segmental and prosodic content recognized from the EMG recordings. The impact will improve the quality of life for people with laryngectomy, and lead to alternative methods for interfacing with computers and machines by people with other speech disorders or limited motor function.
