Assembly workstations are widely used in industrial environments, where pose estimation plays a critical role in ensuring safety and efficiency. In human-robot collaboration, detecting the worker’s pose is essential for preventing collisions and optimizing the interaction between humans and robotic systems. For several years, researchers have been working on pose estimation using either RGB data or point clouds. Point clouds offer a significant advantage over RGB data by directly offering depth information. This paper presents a scalable system that combines low-resolution point clouds from multiple spatially distributed low-cost sensors using CAN-FD for data transmission and then recognizes the human pose using point-based methods. Our system utilizes multiple low-cost time-of-flight sensors that capture the human from different angles. By combining these varied perspectives, a broader context for pose estimation is achieved. Due to the lower resolution of individual sensors and the multiple viewpoints, smaller point clouds must be processed. This enables the use of compact, energy-efficient neural networks. We trained and evaluated our system using reference point cloud data provided by a high-precision stereo vision camera and different persons performing exemplary assembly tasks.