Current Research Projects Currently we have two focused projects. The major funding sources of our reseach are National Institutes of Health and National Science Foundation. Automatic Captioning of Telehealth Conversation Speech This project is on developing a voice-driven captioning system to assist language comprehension of hearing impaired clients in telehealth. Coversation speech between health provider and client is processed online and provider's speech is converted automatically into captions in real-time. In developing such a system, we address research challenges of speech stream separation, spontaneous speech modeling, medicine specialty domain language modeling, fast speaker adaptation, real-time decoding of large vocabulary continuous speech, user interface, etc. Blind Separation and Enhancement of Speech Signal The second project focuses on enhancing speech signals for robust automatic speech recognition and for improving intelligibility and comfort in assistive listening. We address research challenges of blind separation of competing speech sources, enhancement of speech signals from various sources of environmental noises, and far field speech acquisition in reverberant rooms.