Industrial robots are evolving to work closely with humans in shared spaces. Hence, robotic tasks are increasingly shared between humans and robots in collaborative settings. To enable a fluent human robot collaboration, robots need to predict and respond in real-time to worker's intentions. We present a method for early decision using force information. Forces are provided naturally by the user through the manipulation of a shared object in a collaborative task. The proposed algorithm uses a recurrent neural network to recognize operator's intentions. The algorithm is evaluated in terms of action recognition on a force dataset. It excels at detecting intentions when partial data is provided, enabling early detection and facilitating a quick robot reaction.

Depth data has a widespread use since the popularity of high resolution 3D sensors. In multi-view sequences, depth information is used to supplement the color data of each view. This article proposes a joint encoding of multiple depth maps with a unique representation. Color and depth images of each view are segmented independently and combined in an optimal Rate-Distortion fashion. The resulting partitions are projected to a reference view where a coherent hierarchy for the multiple views is built. A Rate-Distortion optimization is applied to obtain the nal segmentation choosing nodes of the hierarchy. The consistent segmentation is used to robustly encode depth maps of multiple views obtaining competitive results with HEVC coding standards.Peer ReviewedPostprint (author's final draft

3D video coding includes the use of multiple color views and depth maps associated to each view. An adequate coding of depth maps should be adapted to the characteristics of depth maps: smooth regions and sharp edges. In this paper a segmentation-based technique is proposed for improving the depth map compression while preserving the main discontinuities that exploits the color-depth similarity of 3D video. An initial coarse depth map segmentation is used to locate the main discontinuities in depth. The resulting partition is improved by fusing a color partition. We assume that the color image is first encoded and available when the associated depth map is encoded, therefore the color partition can be segmented in the decoder without introducing any extra cost. A new segmentation criterion inspired by super-pixels techniques is proposed to obtain the color partition. Initial experimental results show similar compression efficiency to HEVC with a big potential for further improvements.Peer ReviewedPostprint (published version

Multiview color information used jointly with depth maps is a\ud widespread technique for 3D video. Using this depth information,\ud 3D functionalities such as free view point video can be provided\ud by means of depth-image-based rendering techniques. In this pa-\ud per, a new technique to encode depth maps is proposed. Based on\ud the usually smooth structure and the sharp edges of depth map, our\ud proposal segments the depth map into homogeneous regions of ar-\ud bitrary shape and encodes the contents of these regions using dif-\ud ferent texture coding strategies. An optimal lagrangian approach\ud is applied to the hierarchical region representation provided by our\ud segmentation technique. This approach automatically selects the\ud best encoding strategy for each region and the optimal partition to\ud encode the depth map. To avoid the high coding costs of coding\ud the resulting partition, a prediction is made using the associated\ud decoded color imagePeer ReviewedPostprint (published version

In 3D video, view synthesis is used to process new virtual\ud views between encoded camera views. Errors in the coding\ud of the depth maps introduce geometry inconsistencies in\ud synthesized views. In this paper, a 3D plane representation\ud of the scene is presented which improve the performance of\ud current standard video codecs in the view synthesis domain.\ud Depth maps are segmented into regions without sharp edges\ud and represented with a plane in the 3D world scene coordinates.\ud This 3D representation allows an efficient representation\ud while preserving the 3D characteristics of the scene.\ud Experimental results are provided obtaining gains from 10 to\ud 40 % in bitrate compared to HEVC.Peer ReviewedPostprint (published version