In manufacturing shop floors, technologies for worker behavior analysis using video analysis have advanced and are being utilized to enhance productivity. In such video analyses, identifying the individuals is essential. However, in many cases, workers wear the same hats, uniforms, and masks, making individual identification based on facial recognition or clothing difficult. Moreover, to capture the overall scene widely, the recorded videos often show workers as blurred figures. To address these issues, this study proposes a method of individual identification using a graph neural network, taking advantage of a technique that extracts skeleton data through pose estimation. A key characteristic of this method is the aim to simplify annotation work. By combining positioning technology from sensing devices with human tracking in video analysis, we improve the efficiency of building individual identification models. This paper reports an overview of the proposed method, experiments demonstrating its effectiveness, and discussions derived from the experimental results.