Data Collection Methodology Using MediaPipe Framework
The data-collection program has been completed, and the next step is to begin collecting data to construct the dataset. For this purpose, 38 different signs will be collected, with a minimum of 800 data records expected for each sign. The data collection procedure will be executed in the following order: 1.The program will be executed via the terminal with the specified sign name. 2.The individual will present the hand sign to the webcam and commence data collection by pressing ‘0’. The advantage of this data collection program is that the background does not affect the quality of the collected data. This is due to the capability of the MediaPipe framework to recognize the hand regardless of other visual noise in the background. Additionally, the distance between the hand and the camera does not impact data quality, as the program normalizes the joint distances from the wrist joint. To enhance the robustness of the collected data, the following measures will be implemented: ...