Enhancing Control of Microsoft Teams Meetings and PowerPoint Presentations through Gesture Recognition

Improving human-computer interaction is essential for smooth distant collaboration in the age of digital connectivity. In order to expedite user engagements and enhance the overall experience during virtual meetings and presentations, this project explores the integration of gesture recognition technology with Microsoft Teams and PowerPoint. The system uses a custom-built neural network for gesture classification and MediaPipe for hand pose estimation to identify five predefined hand signs: mute, unmute, camera on, camera off, and no action. Using 500 labeled data points (of which 125 were used for testing), the initial testing produced a 100% accuracy score for all motions. These results were further supported by the categorization report, which showed that the F1-score, precision, and recall all achieved 1.00. These outcomes substantiate the resilience of the used neural network architecture, preprocessing methodologies, and data collection strategies. Although the controlled testing environment yielded encouraging results, the study recognizes that additional tuning is required to guarantee resilience in practical applications. Subsequent research endeavors will center around augmenting the dataset, integrating varied settings, and investigating sophisticated methodologies such as data augmentation and transfer learning. To improve the model's accuracy and adaptability over time, adaptive learning techniques and real-time performance monitoring will also be incorporated. This study shows how gesture recognition can be used to build a virtual collaboration environment that is more user-friendly and productive.