Kinect-Based Approach for Upper Body Movement Assessment in Stroke

A review of computational approaches for evaluation of rehabilitation exercises

“…Jung et al [37] proposed a set of distinctive features obtained from experimental data, consisting of mean speed, reaction time, duration, peak velocity, maximum velocity, distance error, direction error, and path length ratio. For assessing upper body movement after stroke, the range of motion, movement speed, symmetry ratio among body sides, and vertical distance were adopted as movement performance indicators in [35]. Yu and Xiong [128] selected eight bone vectors as important features to an algorithm for producing quality scores in support of home-based rehabilitation.…”

“…One shortcoming of the probabilistic models is that the movements are represented at a single level of movement abstraction, and it is difficult to implement probabilistic modeling at multiple levels of movement abstraction. [72] Kinect v1 S Hand-crafted MA Hagler et al [73] Kinect v1 S Hand-crafted MA Nomm and Buhhalko [74] Kinect v1 S Hand-crafted MC Zhao et al [49] Kinect v1 S Hand-crafted MA Antón et al [22] Kinect v1 S Hand-crafted MA Antón et al [55] Kinect v1 S Hand-crafted MC Su [56] Kinect v1 S Hand-crafted MA Su et al [57] Kinect v1 S Hand-crafted MA Crabbe et al [43] (data from [119]) Kinect v1 D None PE Uttarwar and Mishra [75] Kinect v1 S Hand-crafted MC Saraee et al [137] Kinect v2 S Hand-crafted MA Vakanski et al [39] (data from [113]) Kinect v2 S Autoencoder network MA Paiement et al [40] (data from [182]) Kinect v2 S Diffusion maps MC + MA Parisi et al [182] Kinect v2 S Hand-crafted MA Capecci et al [33] Kinect v2 S Hand-crafted MA Capecci et al [34] Kinect v2 S Hand-crafted MA Capecci et al [126] Kinect v2 S Hand-crafted MA Capecci et al [61] Kinect v2 S Hand-crafted MA Capecci et al [62] Kinect v2 S Hand-crafted MA Tao et al [41] (data from [119]) Kinect v2 S Diffusion maps MC + MA Osgouei et al [54] Kinect v2 S Hand-crafted MC Spasojević et al [35] Kinect v2 S Hand-crafted MA Yu and Xiong [128] Kinect v2 S Hand-crafted MA Saraee et al [156] Kinect v2 S Hand-crafted MA Taylor et al [45] Inertial sensor I Hand-crafted MC Zhang et al [48] Inertial sensor I Cross-correlation function MC Lin and Kulić [148] Inertial sensor S None MS Chen et al [183] Inertial sensor I Hand-crafted MC Zhang et al [138] Ine...…”

“…Full-body skeletal data are highly redundant, thus, they are rarely applied directly for modeling and analysis of human motions. Consequently, feature engineering via selection of important skeletal dimensions or distances is often employed for extracting relevant information from skeletal data [33]- [35]. Another common avenue for feature engineering entails creating new local features for representing the motions, based on a set of kinematic parameters that are predefined for an exercise, or by employing a functional mapping [36], [37].…”