This chapter covers the key aspects of Sign Language Recognition (SLR), starting with a brief introduction to the motivations and requirements, followed by a précis of sign linguistics and their impact on the field. The types of data available and the relative merits are explored allowing examination of the features which can be extracted. Classifying the manual aspects of sign (similar to gestures) is then discussed from a tracking and non-tracking viewpoint before summarising some of the approaches to the non-manual aspects of sign languages. Methods for combining the sign classification results into full SLR are given showing the progression towards speech recognition techniques and the further adaptations required for the sign specific case. Finally the current frontiers are discussed and the recent research presented. This covers the task of continuous sign recognition, the work towards true signer independence, how to effectively combine the different modalities of sign, making use of the current linguistic research and adapting to larger more noisy data sets.

We propose a novel hybrid approach to static pose estimation called Connected Poselets. This representation combines the best aspects of part-based and example-based estimation. First detecting poselets extracted from the training data; our method then applies a modified Random Decision Forest to identify Poselet activations. By combining keypoint predictions from poselet activitions within a graphical model, we can infer the marginal distribution over each keypoint without any kinematic constraints. Our approach is demonstrated on a new publicly available dataset with promising results.

Abstract-Human pose estimation in static images has received significant attention recently but the problem remains challenging. Using data acquired from a consumer depth sensor, our method combines a direct regression approach for the estimation of rigid body parts with the extraction of geodesic extrema to find extremities. We show how these approaches are complementary and present a novel approach to combine the results resulting in an improvement over the state-of-the-art. We report and compare our results a new dataset of aligned RGB-D pose sequences which we release as a benchmark for further evaluation.

This research used computer simulation along with human observer studies to evaluate SPECT and planar imaging for lesion detection. The research initially focused on a simple computer simulated phantom consisting of a large sphere with a lesion placed inside at various depths. A constant lesion to background ratio of 8 to 1 was maintained throughout the trials. Two noise levels were simulated by varying the mean value of the sphere and lesion. Both planar and SPECT images were generated simulating a clinical scan time of 30 minutes. Results showed SPECT images consistently had higher contrast but also higher noise values than planar images. When the lesion was located near the center of the sphere the contrast-to-noise ratio was consistently higher in SPECT images. Research using a realistic computer model of the human torso along with realistic image acquisition and reconstruction techniques was then carried out. Planar images were generated with modeling attenuation, scatter and geometric point response. SPECT images were projected and then reconstructed with an OSEM algorithm accounting for attenucation, the scatter and the geometric point response of the collimator.

Robust and fast algorithms for estimating the pose of a human given an image would have a far reaching impact on many fields in and outside of computer vision. We address the problem using depth data that can be captured inexpensively using consumer depth cameras such as the Kinect sensor. To achieve robustness and speed on a small training dataset, we formulate the pose estimation task within a regression and Hough voting framework. Our approach uses random regression forests to predict joint locations from each pixel and accumulate these predictions with Hough voting. The Hough accumulator images are treated as likelihood distributions where maxima correspond to joint location hypotheses. We demonstrate our approach and compare to the state-ofthe-art on a publicly available dataset.

Abstract. Automatic estimation of human pose has long been a goal of computer vision, to which a solution would have a wide range of applications. In this paper we formulate the pose estimation task within a regression and Hough voting framework to predict 2D joint locations from depth data captured by a consumer depth camera. In our approach the offset from each pixel to the location of each joint is predicted directly using random regression forests. The predictions are accumulated in Hough images which are treated as likelihood distributions where maxima correspond to joint location hypotheses. Our approach is evaluated on a publicly available dataset with good results.