Motion tracking and gait feature estimation for recognising Parkinson's disease using MS Kinect

Abstract

Background: Analysis of gait features provides important information during the treatment of neurological disorders, including Parkinson's disease. It is also used to observe the effects of medication and rehabilitation. The methodology presented in this paper enables the detection of selected gait attributes by Microsoft (MS) Kinect image and depth sensors to track movements in three-dimensional space.

Methods: The experimental part of the paper is devoted to the study of three sets of individuals: 18 patients with Parkinson's disease, 18 healthy aged-matched individuals, and 15 students. The methodological part of the paper includes the use of digital signal-processing methods for rejecting gross data-acquisition errors, segmenting video frames, and extracting gait features. The proposed algorithm describes methods for estimating the leg length, normalised average stride length (SL), and gait velocity (GV) of the individuals in the given sets using MS Kinect data.

Results: The main objective of this work involves the recognition of selected gait disorders in both the clinical and everyday settings. The results obtained include an evaluation of leg lengths, with a mean difference of 0.004 m in the complete set of 51 individuals studied, and of the gait features of patients with Parkinson's disease (SL: 0.38 m, GV: 0.61 m/s) and an age-matched reference set (SL: 0.54 m, GV: 0.81 m/s). Combining both features allowed for the use of neural networks to classify and evaluate the selectivity, specificity, and accuracy. The achieved accuracy was 97.2 %, which suggests the potential use of MS Kinect image and depth sensors for these applications.

Conclusions: Discussion points include the possibility of using the MS Kinect sensors as inexpensive replacements for complex multi-camera systems and treadmill walking in gait-feature detection for the recognition of selected gait disorders.

Fig. 1

Data processing presenting a the location of the MS Kinect’s RGB camera and depth sensors, b the flowchart of spatial data acquisition in the given coordinate system, and c fusion of gait parameters to increase the classification accuracy

Fig. 2

An example frame recorded by MS Kinect including a the image frame matrix combined with the skeleton estimate, b the depth frame matrix, c the contour plot of the depth frame matrix with distances from the selected plane, and d the proposed graphical user interface that was used to record MS Kinect data from the observed individuals in a clinical environment and to preview the recorded skeleton, video, and depth sensor data with numbering of joints

Fig. 3

Visualisation of MS Kinect data presenting a the evolution of the z-coordinate of the COM in time with the median values and standard deviations used for the detection of gross errors and outliers rejection, b the relative spatial evolution of the left and right leg centres after the removal of the skeleton mass centre of each frame, c the temporal evolution of the right and left legs movement in three-dimensional space, and d the distances between the leg centres for a selected walk segment of a normal individual

Fig. 4

Results of the evaluation of leg lengths using MS Kinect during gait execution presenting a a histogram of the average leg lengths of separate individuals, b errors in the differences of the lengths of the left and right legs of individuals, and c the normalized stride length distributions for the individuals with Parkinson’s disease (positive set), the age-matched controls (negative set), and the distributions of true and false results across criterion values

Fig. 5

Gait features and their classification into two classes by a the two-layer sigmoidal neural networks 2-4-2 and b the RBF neural networks 36-18-1 with the spread of radial basis functions equal to 0.1

Fig. 6

Selected gait features obtained from MS Kinect, including a the average stride lengths and corresponding standard deviations, b the histograms of the stride length distribution, c the average gait velocities and corresponding standard deviations, and d histograms of the distribution of the velocities for three sets of individuals (i.e. individuals with Parkinson’s disease—PD, age-matched controls—NORM, and the second reference sets of students—STUDENTS )

Fig. 7

The age dependence of a the gait velocity and b the stride length with related regression coefficients

Fig. 8

Sensitivity/specificity plots for the processing of a gait velocity, b stride lengths and classification accuracy presenting results for c the gait velocity features, d the stride length features, and e combined features processing using neural networks