The paper presents two real-time hidden Markov model (HMM)-based systems for recognizing sentence-level continuous American Sign Language (ASL) using a single camera to track the user's unadorned hands. The first system, mounted on a desk, achieves 92% word accuracy, while the second system, worn by the user, achieves 98% accuracy (97% with an unrestricted grammar). Both systems use a 40-word lexicon. The paper discusses the challenges and improvements in ASL recognition, including the use of HMMs, feature extraction, and hand ambiguity handling. The first-person view system, worn by the signer, is shown to be more accurate due to reduced occlusion and better tracking. The authors also discuss future work, including the need for more diverse data, improvements in tracking head and facial gestures, and the potential for adding finger and palm tracking information.The paper presents two real-time hidden Markov model (HMM)-based systems for recognizing sentence-level continuous American Sign Language (ASL) using a single camera to track the user's unadorned hands. The first system, mounted on a desk, achieves 92% word accuracy, while the second system, worn by the user, achieves 98% accuracy (97% with an unrestricted grammar). Both systems use a 40-word lexicon. The paper discusses the challenges and improvements in ASL recognition, including the use of HMMs, feature extraction, and hand ambiguity handling. The first-person view system, worn by the signer, is shown to be more accurate due to reduced occlusion and better tracking. The authors also discuss future work, including the need for more diverse data, improvements in tracking head and facial gestures, and the potential for adding finger and palm tracking information.