A Hidden Semi-Markov Model based approach for rehabilitation exercise assessment

Abstract

In this paper, a Hidden Semi-Markov Model (HSMM) based approach is proposed to evaluate and monitor body motion during a rehabilitation training program. The approach extracts clinically relevant motion features from skeleton joint trajectories, acquired by the RGB-D camera, and provides a score for the subject's performance. The approach combines different aspects of rule and template based methods. The features have been defined by clinicians as exercise descriptors and are then assessed by a HSMM, trained upon an exemplar motion sequence. The reliability of the proposed approach is studied by evaluating its correlation with both a clinical assessment and a Dynamic Time Warping (DTW) algorithm, while healthy and neurological disabled people performed physical exercises. With respect to the discrimination between healthy and pathological conditions, the HSMM based method correlates better with the physician's score than DTW. The study supports the use of HSMMs to assess motor performance providing a quantitative feedback to physiotherapist and patients. This result is particularly appropriate and useful for a remote assessment in the home.

A Deep Learning Framework for Assessing Physical Rehabilitation Exercises

“…This section surveys metrics that have been commonly used in prior studies on rehabilitation assessment [18], [34], [36], [37]. The metrics are classified into two categories: model-less and model-based [9].…”

A review of computational approaches for evaluation of rehabilitation exercises

“…Discrete hidden Markov models (HMM) were implemented for analysis and segmentation of human motion data for rehabilitation exercises [53], [147]. In [148], an approach based on hidden semi-Markov models (HSMM) was applied to evaluate five different rehabilitation exercises and provide an evaluation score [61], [62]. The requirement for segmenting the exercises into individual repetitions by discrete HMM or HSMM was overcome in [40], [41] by employing a continuous HMM.…”

“…Scoring functions are often used to convert the output values of movement evaluation algorithms to a meaningful performance score limited within a certain range [33], [37], [61], [62], [72], [126], [128], [138], [156], [157]. Concretely, it is important that the approaches for movement evaluation generate quality scores that are understandable and interpretable both by patients and medical professionals.…”

“…Although several research works have addressed this problem, the focus has been solely on quantifying rehabilitation movements for static body postures [49] and balance tests [81]. A recent study by Capecci et al [62] proposed rules for assessing rehabilitation exercises based on attained target joint angles, target joint velocities, and postural constraints in accomplishing the goals of each exercise. However, the study proposed guidelines for only five selected rehabilitation exercises.…”

“…The benefit of the distance functions is that they are not exercise-specific, and thus can be applied for evaluation of new types of exercises. Another line of works exploits the ability by probabilistic models with latent variables to encode spatial variability and temporal dynamics of human movements [39]- [41], [54], [60]- [62]. For instance, Gaussian mixture models [39], [63] and hidden Markov models [61], [62] were used for motion modeling, and quality assessment is based on the likelihood that the individual sequences are being drawn from a derived model.…”