System for automatic gait analysis based on a single RGB-D camera

Abstract

Human gait analysis provides valuable information regarding the way of walking of a given subject. Low-cost RGB-D cameras, such as the Microsoft Kinect, are able to estimate the 3-D position of several body joints without requiring the use of markers. This 3-D information can be used to perform objective gait analysis in an affordable, portable, and non-intrusive way. In this contribution, we present a system for fully automatic gait analysis using a single RGB-D camera, namely the second version of the Kinect. Our system does not require any manual intervention (except for starting/stopping the data acquisition), since it firstly recognizes whether the subject is walking or not, and identifies the different gait cycles only when walking is detected. For each gait cycle, it then computes several gait parameters, which can provide useful information in various contexts, such as sports, healthcare, and biometric identification. The activity recognition is performed by a predictive model that distinguishes between three activities (walking, standing and marching), and between two postures of the subject (facing the sensor, and facing away from it). The model was built using a multilayer perceptron algorithm and several measures extracted from 3-D joint data, achieving an overall accuracy and F1 score of 98%. For gait cycle detection, we implemented an algorithm that estimates the instants corresponding to left and right heel strikes, relying on the distance between ankles, and the velocity of left and right ankles. The algorithm achieved errors for heel strike instant and stride duration estimation of 15 ± 25 ms and 1 ± 29 ms (walking towards the sensor), and 12 ± 23 ms and 2 ± 24 ms (walking away from the sensor). Our gait cycle detection solution can be used with any other RGB-D camera that provides the 3-D position of the main body joints.

Validation of a Single RGB-D Camera for Gait Assessment of Polyneuropathy Patients

“…The experimental setup, illustrated in Figure 1, consisted of a reference motion capture system (Qualisys) that included twelve Oqus infrared cameras and 61 retro-reflective markers attached to the patients’ body as indicated in [10], according to the setup typically used in LABIOMEP for full body gait analysis. The setup also comprised a single RGB-D camera (Kinect v2) placed in front of the participant, with a height of 1 m and tilt angle of −5 degrees, to maximize the sensor’s practical depth range.…”

“…Data provided by the Kinect—3-D body joint, depth and infrared—were acquired at 30 Hz, using our KiT software application [25]. Each body joint frame includes the 3-D position of the twenty-three joints tracked by Kinect (see [10]). The 3-D coordinate system associated with the Kinect v2 is illustrated in Figure 2.…”

“…For the RGB-D camera, to decide if a given frame corresponds to a heel strike instant, we relied on the distance between ankles signal and a window with a size of ND1 frames centered at the current frame, as further detailed in [10]. The side (left or right) associated with each detected heel strike was identified relying on the ankles’ velocity signals, considering a window with a size of ND2 frames centered at the corresponding heel strike instant frame [10]. The used signals were previously processed using a moving average filter, considering a window size of NF1 and NF2 frames for heel strike detection and side identification, respectively.…”

“…The Kinect has already been validated in different contexts for both healthy subjects [16,17,18,19,20] and patients with movement impairments, such as the assessment of the maximum walking speed and static posturography in multiple sclerosis [21,22,23] and the assessment of different movements in Parkinson’s Disease [24]. In the case of gait analysis, it was found that the Kinect can be used for obtaining different spatiotemporal gait parameters/measures in healthy populations [9,10,18,19] and patients with multiple sclerosis [21]. However, its use is not recommended for measuring most kinematic gait parameters/measures, in healthy populations [9,19,20].…”

“…Instead of exploring the time series for the whole gait trial as in [9], we computed 23 spatiotemporal and kinematic parameters for each performed gait cycle. These gait parameters include ten more spatiotemporal and eight kinematic parameters that were not considered in our previous work on automated gait analysis [10]. To compute some of the parameters, it was further necessary to detect a new type of gait event (toe off).…”