This SBIR Phase I project aims to create an emotionally expressive software-based speech-generating communication system, which is designed for individuals with little or no verbal communication ability -- particularly those with autism spectrum disorder. Current alternative communication technology does not offer the ability to express emotional content: something that is crucial to effective, comprehensible communication at home, in the community, and at work. The ability to convey emotions -- sadness, anger, or happiness, for example -- through one's tone of voice or facial expressions has far-reaching educational and vocational benefits for individuals with communication challenges. In other words, emotional content helps clarify the communicative intent behind a spoken message. When non-verbal individuals can communicate thoroughly and effectively, they increase their ability to receive a free and appropriate public education, and they expand their vocational opportunities. When individuals can receive an education and find a job, they can contribute back to society. The proposed work supports progress in science and engineering, yet it also enhances the potential for current educational applications and future research studies. Finally, this project builds on Federal and State efforts to offer people with disabilities educational and vocational services.
This project will develop and combine novel software algorithms that target digital facial recognition patterns and speech-based waveforms, in order to enhance the emotional content of conversational communication for individuals who cannot speak volitionally. The ability to add emotional content onto synthesized speech is a new technology that could substantially benefit individuals with communication challenges, such as people with autism spectrum disorder. This project seeks to facilitate and augment emotional expression and, therefore, increase communicative competence for individuals who use software-based synthesized speech systems to express their ideas, feelings, and wishes. Measuring whether or not this project accurately conveys human emotion in speech and facial expressions requires a perceptual test involving neurotypical adults. Specifically, the perceptual test will include both trained and untrained judges. The purpose of this is to ensure that the facial expressions as well as the pitch contour and vocal emphasis align with the emotion labels (i.e., happy, sad, or mad). In order to accurately calculate the judges' inter-rater reliability, the study will use a method that determines the extent to which judges' ratings agree, relative to how much they would likely conform, if they were to randomly rate the same stimuli.