A real-time speech-to-text personal computer dictation program has recently been marketed with an active vocabulary in excess of 22,000 words. Although fast and accurate, the program requires each speaker to register his or her speech pronunciation through a tedious process of word repetition, allowing the computer to recognize each individual's word enunciation. While acceptable for the intended purpose of personal dictation, the program is inadequate for use in a classroom or training environment, where hearing-impaired students require speech-to text translation for a continually changing group of teachers and peers. To create an autonomous real-time speech-to-text translation system, the proposed SBIR effort will develop a voice processing circuit to electronically compress the wide range of frequencies associated with human speech into a narrow frequency bandwidth, creating a standard synthetic voice which is identical for every speaker. The speech-to-text dictation program can be trained to recognize word pronunciations provided by the monotone synthetic voice, allowing multiple speakers to use the system without a lengthy voice registration process. Transcribed in real-time, spoken words can be displayed on the terminal, projected to a large screen, or stored to file for later retrieval.Anticipated Benefits/Potential Commercial Applications of the Research:Phase I will determine the feasibility of using an electronic voice processing circuit to provide uniform speech input for a real-time, speech-to-text conversion program, and will determine methods to effectively integrate speech-to-text computer systems into classroom and training program environments. The voice processing circuit will be further tested and refined in Phase II, leading to the development of a truly autonomous speech-to text translation system. The ability to provide real-time speech-to-text translation without the need for lengthy voice registration is an innovation with myriad commercial applications. The autonomous speech-to-text conversion system could find widespread use in elementary, high-school, and postsecondary classrooms and training programs, allowing deaf students to more effectively communicate with their teachers and peers. The technology may be incorporated into notebook computers, providing a portable real-time translation system for personal use by hearing-impaired individuals. Beyond the classroom, voice processed speech-to-text technology may find use in general business offices, not only as an aid to hearing-impaired professionals but also as an economic method to provide immediate, multiple user access to a single computer dictation system.
