Human-computer interaction (HCI) methods that are more intuitive and accessible continue to be in high demand. Conventional interfaces have drawbacks, particularly for people with disabilities or in particular settings. A real-time hand gesture recognition system is presented in this study designed to address these challenges. Our system combines two complementary methods: keypoint classification for detecting static hand poses, and point history classification for recognizing dynamic gestures. These two approaches create a versatile system capable of interpreting a wider range of user commands. Custom datasets were compiled for both model types and trained independently using neural network architectures tailored to the data. We present the implemented system architecture, model evaluation, and potential applications with a focus on assistive technologies.