Using Blind Source Separation on accelerometry data to analyze and distinguish the toe walking gait from normal gait in ITW children

Application of machine learning to the identification of joint degrees of freedom involved in abnormal movement during upper limb prosthesis use

To evaluate movement quality of upper limb (UL) prosthesis users, performance-based outcome measures have been developed that examine the normalcy of movement as compared to a person with a sound, intact hand. However, the broad definition of “normal movement” and the subjective nature of scoring can make it difficult to know which areas of the body to evaluate, and the expected magnitude of deviation from normative movement. To provide a more robust approach to characterizing movement differences, the goals of this work are to identify degrees of freedom (DOFs) that will inform abnormal movement for several tasks using unsupervised machine learning (clustering methods) and elucidate the variations in movement approach across two upper-limb prosthesis devices with varying DOFs as compared to healthy controls. 24 participants with no UL disability or impairment were recruited for this study and trained on the use of a body-powered bypass (n = 6) or the DEKA limb bypass (n = 6) prosthetic devices or included as normative controls. 3D motion capture data were collected from all participants as they performed the Jebsen-Taylor Hand Function Test (JHFT) and targeted Box and Blocks Test (tBBT). Range of Motion, peak angle, angular path length, mean angle, peak angular velocity, and number of zero crossings were calculated from joint angle data for the right/left elbows, right/left shoulders, torso, and neck and fed into a K-means clustering algorithm. Results show right shoulder and torso DOFs to be most informative in distinguishing between bypass user and norm group movement. The JHFT page turning task and the seated tBBT elicit movements from bypass users that are most distinctive from the norm group. Results can be used to inform the development of movement quality scoring methodology for UL performance-based outcome measures. Identifying tasks across two different devices with known variations in movement can inform the best tasks to perform in a rehabilitation setting that challenge the prosthesis user’s ability to achieve normative movement.

Identification of Diseases Based on the Use of Inertial Sensors: A Systematic Review

Inertial sensors are commonly embedded in several devices, including smartphones, and other specific devices. This type of sensors may be used for different purposes, including the recognition of different diseases. Several studies are focused on the use of accelerometer signals for the automatic recognition of different diseases, and it may empower the different treatments with the use of less invasive and painful techniques for patients. This paper aims to provide a systematic review of the studies available in the literature for the automatic recognition of different diseases by exploiting accelerometer sensors. The most reliably detectable disease using accelerometer sensors, available in 54% of the analyzed studies, is the Parkinson’s disease. The machine learning methods implemented for the automatic recognition of Parkinson’s disease reported an accuracy of 94%. The recognition of other diseases is investigated in a few other papers, and it appears to be the target of further analysis in the future.

Computing the Partial Correlation of ICA Models for Non-Gaussian Graph Signal Processing

Conventional partial correlation coefficients (PCC) were extended to the non-Gaussian case, in particular to independent component analysis (ICA) models of the observed multivariate samples. Thus, the usual methods that define the pairwise connections of a graph from the precision matrix were correspondingly extended. The basic concept involved replacing the implicit linear estimation of conventional PCC with a nonlinear estimation (conditional mean) assuming ICA. Thus, it is better eliminated the correlation between a given pair of nodes induced by the rest of nodes, and hence the specific connectivity weights can be better estimated. Some synthetic and real data examples illustrate the approach in a graph signal processing context.

The utility of normative foot floor angle data in assessing toe-walking

Initial heel contact is an important attribute of gait, and failure to complete the heel rocker reduces gait stability. One common goal in treating toe-walking is to restore heel strike and prevent or reduce early heel rise. Foot floor angle (FFA) is a measure of toe-walking that is valuable for quantifying foot orientation at initial contact when using ankle dorsiflexion angle alone is misleading. However, no age-standardized FFA norms exist for clinical evaluation. Our objectives were to: (1) obtain normative FFA in typically developing children; and (2) examine its utility in the example of toe-walking secondary to unilateral cerebral palsy. Gait kinematics were acquired and FFA trajectories computed for 80 typically developing children (4-18 years). They were also obtained retrospectively from 11 children with toe-walking secondary to unilateral cerebral palsy (4-10 years), before and after operative intervention, and compared to 40 age-matched, typically developing children. FFA at initial contact was significantly different (P<.001) between pre-surgery toe-walking (-14.7±9.7°; mean±standard deviation) and typical gait (18.7±2.8°). Following operative lengthening of the gastrocnemius-soleus complex on the affected side, FFA at initial contact (-0.9±5.3°) was significantly improved (P<.001). Furthermore, several cases were identified for which the sole use of ankle dorsiflexion angle to capture toe-walking is misleading. The assessment of FFA is a simple method for providing valuable quantitative information to clinicians regarding foot orientation during gait. The demonstrated limitations of using ankle dorsiflexion angle alone to estimate foot orientation further emphasize the utility of FFA in assessing toe-walking.

Double-layer ensemble monitoring of non-gaussian processes using modified independent component analysis

The modified independent component analysis (MICA) was proposed mainly to obtain a consistent solution that cannot be ensured in the original ICA algorithm and has been widely investigated in multivariate statistical process monitoring (MSPM). Within the MICA-based non-Gaussian process monitoring circle, there are two main problems, i.e., the selection of a proper non-quadratic function for measuring non-Gaussianity and the determination of dominant ICs for constructing latent subspace, have not been well attempted so far. Given that the MICA method as well as other MSPM approaches are usually implemented in an unsupervised manner, the two problems are always solved by some empirical criteria without respect to enhancing fault detectability. The current work aims to address the challenging issues involved in the MICA-based approach and propose a double-layer ensemble monitoring method based on MICA (abbreviated as DEMICA) for non-Gaussian processes. Instead of proposing an approach for selecting a proper non-quadratic function and determining the dominant ICs, the DEMICA method combines all possible base MICA models developed with different non-quadratic functions and different sets of dominant ICs into an ensemble, and a double-layer Bayesian inference is formulated as a decision fusion method to form a unique monitoring index for online fault detection. The effectiveness of the proposed approach is then validated on two systems, and the achieved results clearly demonstrate its superior proficiency.

Performance Analysis of ICA in Sensor Array

As the best-known scheme in the field of Blind Source Separation (BSS), Independent Component Analysis (ICA) has been intensively used in various domains, including biomedical and acoustics applications, cooperative or non-cooperative communication, etc. While sensor arrays are involved in most of the applications, the influence on the performance of ICA of practical factors therein has not been sufficiently investigated yet. In this manuscript, the issue is researched by taking the typical antenna array as an illustrative example. Factors taken into consideration include the environment noise level, the properties of the array and that of the radiators. We analyze the analytic relationship between the noise variance, the source variance, the condition number of the mixing matrix and the optimal signal to interference-plus-noise ratio, as well as the relationship between the singularity of the mixing matrix and practical factors concerned. The situations where the mixing process turns (nearly) singular have been paid special attention to, since such circumstances are critical in applications. Results and conclusions obtained should be instructive when applying ICA algorithms on mixtures from sensor arrays. Moreover, an effective countermeasure against the cases of singular mixtures has been proposed, on the basis of previous analysis. Experiments validating the theoretical conclusions as well as the effectiveness of the proposed scheme have been included.