Database Open Access

KINECAL

Sean Maudsley-Barton Moi Hoon Yap

Published: June 8, 2023. Version: 1.0.3


When using this resource, please cite: (show more options)
Maudsley-Barton, S., & Yap, M. H. (2023). KINECAL (version 1.0.3). PhysioNet. https://doi.org/10.13026/vvkp-ct80.

Please include the standard citation for PhysioNet: (show more options)
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.

Abstract

The field of human action recognition has made great strides in recent years, much helped by the availability of a wide variety of datasets that use Kinect to record human movement. Conversely, progress towards using Kinect in clinical practice has been hampered by the lack of appropriate data. In particular, datasets that contain clinically significant movements and appropriate metadata. This paper proposes a dataset to address this issue, namely KINECAL. It contains the recordings of 90 individuals carrying out 11 movements, commonly used in the clinical assessment of balance. The dataset contains relevant metadata, including clinical labelling, falls history and postural sway metrics. In addition, the recordings were made in both the lab and in informal settings.


Background

Complications associated with injurious falls are the most common cause of death for those aged over 65. In the UK, 1/3 of those aged over 65 and 1/2 of those aged 80 will fall once a year. In addition to the immediate pain and discomfort, even a single fall can lead to a range of associated conditions such as depression, anxiety and social isolation. In order to identify those in need of help, a falls-risk assessment must be carried out. However, there is a dichotomy at the heart of falls-risk assessment. Lab-based research tends to use expensive equipment (force plates and marker-based motion capture) to quantify balance impairment. Even here, the costs mean that force plate data tends to dominate.

In clinical practice, observational tests are the dominant form of assessment. This means there is difficulty in translating state of the art research into clinical practice. In addition, clinical tests which have shown great utility in falls risk assessment, such as the Sit to Stand x 5 (STS-5) and 3m walk, are difficult to instrument using a force plate. These tests could be instrumented using marker-based motion capture, but even if one ignores the cost of such systems, the space required and set-up time make this option impractical for everyday assessment.

Markerless motion capture could provide a practical solution to bridge the gap between research and practice. Markerless solutions can provide joint angles akin to those derived from marker-based solutions and sway metrics akin to those derived from force plates. These systems are not without issues, but they can provide insight, difficult to achieve any other way, away from the lab. However, research into their use as an objective method of assessing balance, frailty and falls risk is under-researched. As is often the case, there is a chicken and egg situation when it comes to appropriate data on which to base the development of useful methods of assessment using these devices. The proposed dataset, named KINECAL, addresses this need.

Kinect, and more generally Red Green Blue + Depth (RGB+D) data, is used extensively in HAR research. This area of research has made huge strides in recent years, driven forward by the availability of a diverse range of publicly available datasets. On the face of it, it would seem this type of dataset would be useful in the clinical study of human movement. However, the focus is quite different. HAR aims to identify a small set of human movements (actions) from a nearly infinite set of possibilities. The clinical use of motion capture seeks to quantify the quality of a prescribed set of movements. In addition, anonymised clinically-important metadata such as age, height, weight, and meaningful labels of impairment are rarely included in HAR datasets.

In his review of RGB+D datasets, Firman detailed 45 datasets, 44 of which were recorded using Kinect. Of these datasets, only one, the K3Da dataset [1] had movements that are useful in the assessment of balance disorders, it has some basic metadata (age, height, weight), but it has no labels of impairment or falls history. Since the publication of the Firman paper, a few datasets have been created to address these issues. The Multimodal Dataset [2] contains clinically relevant movements (Timed get-up-and-go (TUG), a single 30 second chair stand, a 45 second unilateral stance, and a 2-minute step test). However, the dataset comprises just 21 subjects, evenly split between young and old, and none of the participants had known movement impairment at the time of testing.

The KIMORE Dataset [3] has 78 subjects, 44 healthy mean age 36.7 ± 16.8 years and 32 suffering from motor dysfunction mean age 60.44 ± 14.2 years. The dataset includes clinical scoring of the movements. However, the movements do not relate directly to falls-risk. Table 1 provides a comparison between KINECAL and other datasets containing clinical movements.

Table 1: A comparison of KINECAL and previous datasets that contain clinically significant movements.
  Multimodal Dataset K3Da KIMORE KINECAL
  [1] [2] [3]  
Number of
participants
21 54 78 90
Depth data No Yes Yes Yes
Skeleton data Yes Yes Yes Yes
Clinical
evaluation
No No Yes Yes
Falls related
labels
No No No Yes
Postural Sway
Metrics
No No No Yes
Recordings in
informal settings
No No No Yes

Methods

Participants

90 participants, aged from 18 to 92, were recorded for the KINECAL dataset, carrying out a range of movements commonly used in clinical tests. The recordings were made using a Kinect V2. Further details of data collection are shown below. Participants were excluded if they had any of the following:

  • Treatment for cancer in the previous 2 years
  • Joint replacement in the previous year
  • Broken a leg or hip bone or had a joint replaced, e.g. hip or knee, in the previous 2 years
  • Any lower limb amputation
  • Suffering from neuromuscular conditions (e.g. multiple sclerosis)
  • Have been diagnosed with Alzheimer’s or Dementia
  • Cannot read and communicate in English, either verbal or written.

Clinical significant movements

Table 2 provides a list of the clinically significant movements included in the dataset, along with the instructions given to each participant. The movements are: Quiet standing, eyes open (EO) and eyes closed (EC), semi-tandem stance EO, tandem stance EO, 3m walk, and 5 times sit to stand (this constitutes the Short Physical Performance Battery (SPPB) assessment).

To assess a participant's ability to react to perturbation, the quiet standing trials EO/EC were repeated while standing on a compliant surface (Airex balance pad). Finally, two movements that are often used as standalone tests are included, i.e. TUG and unilateral stance EO/EC.

Table 2: A description of movements in the KINECAL and how they were described to the participants.
  Standard instruction to the participants
STS-5 From a seated position, with your arms crossed over
your chest, rise, extending your legs fully, then sit
down again. Repeat five times, as quickly as possible.
Quiet standing,
Eyes open, Firm surface
Stand feet close together, eyes open, and arms by your side.
The test is terminated after 20 seconds or if lost balance.
Quiet standing,
Eyes closed, Firm surface
Stand feet close together, eyes closed and arms by your side.
The test is terminated after 20 seconds or if lost balance.
Quiet standing,
Eyes open, Foam
Same instructions as for standing on a firm surface,
but perform on a foam.
Quiet standing,
Eyes closed, Foam
Same instructions as for standing on a firm surface,
but perform on a foam.
Semi-tandem Stance Stand with the toe of the back foot against the side of
the heal of the front foot.
Ether foot can be forward, whichever is most comfortable.
The test is terminated after 20 seconds.
Tandem Stance Stand with the toe of the back foot against the
back of the heal of the front foot.
Ether foot can be forward, whichever is most comfortable
The test is terminated after 20 seconds.
Unilateral stance,
Eyes open
Stand on one leg, whichever is most comfortable, with the other leg
flexed 6 inches off the ground, hands by your side, eyes open.
The test is terminated after 20 seconds or when the
lifted leg touches the ground.
Unilateral stance,
Eyes closed
Stand on one leg, whichever is most comfortable, with the other leg
flexed 6 inches off the ground, hands by your side, eyes closed.
The test is terminated after 20 seconds or when the
lifted leg touches the ground.
TUG From a seated position, stand, and walk to a marker
3m away and return to the seat
3m walk From a standing position, walk to a marker
3m away, and return to the seat

The participants are grouped by both age and falls-risk. This enables researchers to disambiguate factors that relate to each.

Data Collection

Recordings were made using a Kinect camera mounted on a tripod at a height of 1.14m from the ground plane. Participants were asked to stand 3m from the camera for static stances and 4m for TUG and 3m walk (the additional meter was needed to capture the entire 3m walk). Participants were recorded using custom software (skel recorder), written using Visual Studio 2015 and the Kinect SDK 2.0. It captures Depth and Skeleton data and stores it on a hard disk. Recodings were made in laboratory conditions and more informal settings, such as church halls.

Dataset structure

In designing KINECAL, we reviewed previous studies relating to aging and falls-risk. Some studies have considered just age-related changes, e.g. young (<35 years old) vs older (>65 years old) populations. This polarisation is helpful if the question is, "how does postural sway change with age". However, using this type of data alone can be problematic when applied to falls-risk. While aging plays a part in falls-risk, it is not the whole story. For example, if we consider master athletes (individuals over the age of 30, but many of whom are aged over 80, who still participate in athletic competition).

In spite of demanding regimes of training and competition, the occurrence of falls for master athletes was found to be no higher than it was for younger athletes. While they are seen as functionally fitter, compared to those of a similar age who do not compete in sporting events, the effects of aging are still detectable as changes in motor neurons, but this does not translate to an increase in the number of falls. McPhee et al. suggested that the protective effects of regular exercise are universal and not limited to elite athletes.

An alternative approach, when considering falls-risk, is to control for age. This is the approach taken by [5,6], who compared aged-matched populations of older fallers and non-fallers. However, this approach can make it just as challenging to understand the interaction of age and physical function. With these factors in mind, we structured KINECAL to help aid the separation of falls-risk from age effects. The following section details how participants were labelled to provide a range of groupings when working with the data.

Self-reported labels

All participants were asked the following question. "Have you had any fall, including a slip or trip in which you lost your balance and landed on the floor or ground or lower level in the past 12 months?" Possible answers were [ None, One, Two, Three, Four or more]. Based on the answer they gave and their age, participants were split into the following groups:

  • Healthy-Adult: members of this group, were all < 65 years old and gave the answer None
  • Non-Faller: members of this group, were all > 65 years old and gave the answer None
  • Self-reported-Faller: members of this group were all > 65 years old and answered One, Two, Three, Four or more.

Table 3, details the numbers in each group, along with their age range and the gender split. These groups are exclusive.

Table 3: Grouping of Self-reported Labels: This table shows the split between the different groups in terms of mean age, total numbers and gender split. A description of each group is also included.
  Description Age (± 95% CI) Total
Number
Male Female
Healthy-Adult < 65 years,
no history of falls
in the last 12 months
mean age 46.2 (±22.7) 33 22 11
Non-Faller >= 65 years,
no history of falls
in the last 12 months
mean age 73.3 (±11.7) 33 16 17
Self-Reported-Faller >= 65 years,
reported >= 1 falls
in the last 12 months
mean age 72.6 (±13.6) 24 15 9

Single and multiple fallers

The Self-reported-Faller group can be further split into single and multiple fallers. We make this distinction because someone who answered "One" to the falls-history question might simply be the victim of bad luck and not someone with a high likelihood of future falls. Someone who declared they had fallen multiple times is more likely to suffer future falls and could be thought of as a "true faller" It is up to the individual researcher to make the best use of these labels.

  • Self-reported-Faller_s: members of this group were all > 65 years old and gave the answer "One".
  • Self-reported-Faller_m: members of this group were all > 65 years old and answered "Two", "Three", or "Four or more".

Table 4, details the numbers in each group, along with their age range and the gender split. Note: the Self-reported-Faller_s and Self-reported-Faller_m groups are made up of members of the Self-Reported-Faller group. Participants are multiple-labelled, which provides flexibility in use. However, it is up to the individual researcher to decide how to best use these labels.

Table 4: Grouped Sub-labelling of Self-reported-Fallers: This table shows the split between the different single (_s) and multiple fallers (_m) in terms of mean age, total numbers and gender split. A description of each group is also included.
  Description Age (±95% CI) Total
Number
Male Female
Self-reported-Faller_s >= 65 years,
reported 1 fall,
in the last 12 months
mean age 72.3 (±15) 15 7 8
Self-reported-Faller_m >= 65 years,
reported >1 fall,
in last 12 months
mean age 73.1 (±11.7) 9 2 7

Clinical labelling

The movements recorded in the dataset provide an alternative means of labelling participants. Moreover, one which relates to clinical tests of physical impairment, hence we call this clinical labelling. Clinical labelling was a two-stage process. 1) The recordings for each participant were played back and marked against accepted thresholds for physical impairment and falls-risk for each of the following tests: SPPB; Slow 3m walk; TUG and slow time to complete STS-5. 2) Any individual categorised as impaired for a least two of these tests was labelled Clinically-At-Risk in the dataset. Using this method, six people were identified as belonging to the Clinically-At-Risk group.

Details of this group are shown in Table 5. The thresholds of impairment for each test are discussed below. Note: the Clinically-At-Risk group is made up of members of the Non-Faller group (2) and Self-Reported-Faller group (4). Participants are multipally labelled, which provides flexibility in use. However, it is up to the individual researcher to decide how to best use these labels.

Table 5: Details of the Clinically-At-Risk group: mean age, total numbers and gender split are shown
  Description Age (±95% CI) Total
Number
Male Female
Clinically-At-Risk >= 65 years
identified as impaired
>= 2 clinical tests
mean age 80.3 (±11.8) 6 1 5

Thresholds of clinical impairment

This section details the thresholds applied to each clinical test.

SPPB

The SPPB test was carried out and scored using the protocol described in [7]. Based on the SPPB score, the participant was classified using the scheme from and detailed in Table 6. Those classified as having Moderate Limitations or Severe Limitations were marked as at-risk for this test.

Table 6: SPPB classification
Score Classification
0-3 Severe Limitations
4-6 Moderate Limitations
7-9 Mild Limitations
10-12 Normal

3m walk

Quach et al. used a 3m walk to assess falls-risk in an 18-month-long, longitudinal study of 764 community-dwelling older people, mean age: 78 ± 5 years [8]. They concluded that a walking speed of < 0.6 m/s was associated with an increased risk of falling inside. This is equivalent to a time of > 5 seconds to complete. Applying this threshold to the 3m walks in KINECAL, anyone who took >5 seconds to complete the 3m walk was marked as at-risk for this test.

STS-5

Another longitudinal study was undertaken by Ward et al [9]. Over 4 years, they studied 755 community-dwelling older people, mean age: 78.1 ± 5.4 years. The study looked at SPPB as a predictor of injurious falls. The conclusion was that the STS-5 alone was enough to assess falls-risk. They suggested that a time to complete > 16.7 seconds may be sufficient to identify those at risk of future falls. Applying this threshold to KINECAL, anyone who took >16.7 seconds to complete the STS-5 test was marked as at-risk, for this test.

Timed “Up & Go”

Shumway-Cook et al [4, 10]. found Timed “Up & Go” (TUG) to be a powerful test for fallers. They used a population of 30 people, split evenly between two groups, i.e. Non-Fallers (no historic falls in the last 6 months) with mean age: 78 ± 6 years, and Fallers ( > 2 falls in the last 6 months by history) with mean age: 86.2 ± 6 years. They suggested that those who take > 14 seconds to complete the TUG are at elevated risk of falls. The Mc Kinly Laboratory provides a reference for normative scores for TUG, synthesised from the Shumway-Cook paper and 3 more. This reference extends the recommendations of Shumway-Cook et al., as presented in Table 7. In spite of the other work, the value presented in [4] remains the current recommendation to identify those at elevated risk of falls. Applying this threshold to KINECAL, anyone who took >14 seconds to complete the STS-5 test was marked as at-risk for this test.

Table 7: TUG classification
Time Classification
30 seconds Problems, cannot go outside alone, requires gait aid
20 seconds Good mobility can go out alone, mobile without a gait aid
14 seconds Elevated risk of falls
10 seconds Normal

Filtering

As with any system that digitises real-world data, Kinect¢s signal is bound up with noise, which can be seen as jittering of the skeletal joints. To address this issue, the recordings were filtered using a Butterworth fourth-order zero-lag filter, with a cut-off frequency of 8 Hz. The cut-off frequency was calculated as described by Winter.

Pose normalisation

Each skeleton frame was aligned to the first frame of the recording, making all subsequent movements relative to this initial position, using equation 1.

p n , i ( x , y , z ) = P n , i ( x , y , z ) P 0 , S P I N E _ B A S E ( x , y , z ) ( 1 ) p_{n,i}(x,y,z)^*= P_{n,i}(x,y,z) - P_{0,{SPINE\_BASE}}(x,y,z) (1)

Where pn, i(x,y,z)* represents the normalised position of the x,y, z-axis of joint i in frame n. P0, SPINE_BASE(x,y,z) represents the position of the SPINE_BASE joint in the first frame of the recording, and Pn, i(x,y,z) represents the positions of joint i in frame n.

Estimation of center of mass

In posturography, postural sway relates to the change in the position of the body¢s Center of Mass (CoM) over time. The most commonly used device for posturography is a force plate. When using force plates, the CoM position is estimated from the Centre of Pressure (CoP) at the surface of a force plate. Using the inverted-pendulum model, the position of the CoM is related to the position of the CoM by applying an offset related to an individual's height. In our work, we derive CoM using a method first described by Leightley et al. It calculates CoM as the 3D Euclidean mean of 3 joints of the Kinect skeleton: Hip left, Hip right, Spine mid, as described in equation 2. This method has been validated to be equivalent to the CoM derived from the Balance Master in a previous paper.

C o M x = H I P _ L E F T x + H I P _ R I G H T x + S P I N E _ M I D x 3 C o M y = H I P _ L E F T y + H I P _ R I G H T y + S P I N E _ M I D y 3 C o M y = H I P _ L E F T z + H I P _ R I G H T z + S P I N E _ M I D z 3 C o M = [ C o M x , C o M y , C o M z ] ( 2 ) \begin{equation} \label{eq:KINECAL_CoM} \begin{split} &CoM_x=\frac{HIP\_LEFT_x+ HIP\_RIGHT_x + SPINE\_MID_x}{3} \\ \\ &CoM_y=\frac{HIP\_LEFT_y+ HIP\_RIGHT_y + SPINE\_MID_y}{3} \\ \\ &CoM_y=\frac{HIP\_LEFT_z+ HIP\_RIGHT_z + SPINE\_MID_z}{3} \\ \\ \\ &CoM = [CoM_x, CoM_y, CoM_z] \end{split} \end{equation} (2)

Where HIP_LEFT, HIP_RIGHT, SPINE_MID are Kinect joints. The Figure also shows the resultant position of the CoM for a single frame.

Generation of sway metrics

The 3D CoM time series is a useful asset for researchers, as is the full depth and skeleton data. However, a set of common force plate metrics has been included for all upright stances. In doing this, the dataset simulates the type of lab-based metrics you would obtain from a force plate. Hence in one dataset, you have clinical and lab-based tests for each participant. Generally, when working with sway metrics derived from force plates, the CoM position is expressed in only two dimensions, i.e. the anatomical directions Anterior-Posterior (AP) and Mediolateral (ML). Kinect uses a right-handed worldview the origin is the centre of the camera. This means the z-axis of Kinect was mapped to movement in the AP direction of the person being recorded. Similarly, the x-axis was mapped to movement in the ML direction. For this application, the y-axis was ignored. Table 8, shows the abbreviations used for each metric, alongside a description of each metric. The abbreviations follow the form laid out in Prieto et al.

Table 8: This table details the sway metrics from the KINECAL dataset. RD refers to the Resultant Distance, AP refers to Anterior-posterior and ML refers to Medio-lateral directions, details of how these metrics were calculated can be in the "Calculation of sway metrics" section
  Description
MDIST Mean distance of the RD time series
RDIST RMS distance of the RD time series
MVELO Mean velocity of the RD time series
TOTEX Total excursions of the RD time series
MFREQ Mean frequency of the RD time series
   
MDIST_AP Mean distance of the AP time series
RDIST_AP RMS distance of the AP time series
MVELO_AP Mean velocity of the AP time series
TOTEX_AP Total excursions of the AP time series
MFREQ_AP Mean frequency of the AP time series
   
MDIST_ML Mean distance of the ML time series
RDIST_ML RMS distance of the ML time series
MVELO_ML Mean velocity of the ML time series
TOTEX_ML Total excursions of the ML time series
MFREQ_ML Mean frequency of the ML time series
   
AREA_CE 95% Confidence Area

Time series

Force plates often have marks on their surface proscribing foot positions or will atomically centre the recordings. In the case of the recordings presented here. Centring was achieved by subtracting the mean CoM position, for an entire recording, from the CoM value at each time step. This was done separately for movement in the AP and ML directions using equations (3). The mean position was calculated using equations (2). The output of equations (3) was concatenated to produce the AP and ML time series. A third-time series was calculated, which takes into account movements in the AP and ML directions in a single value. Known as the resultant distance time series (RD), it was calculated as the vector distance from the mean CoM position to a pair of points in the AP and ML time series at each time step. Values for each time step were generated using equation (4) and then concatenated to produce RD. Each of the calculated sway metrics was calculated using all three time series.

M L _ r a w i ¯ = 1 n i = i n M L _ r a w i A P _ r a w i ¯ = 1 n i = i n A P _ r a w i ( 3 ) \begin{equation} \label{eq:KINECAL_mean} \begin{split} \bar{ML\_raw_i} = \frac{1}{n} \sum_{i=i}^{n} ML\_raw_{i}\\ \\ \bar{AP\_raw_i} = \frac{1}{n} \sum_{i=i}^{n} AP\_raw_{i} \end{split} \end{equation} (3)
A P i = A P _ r a w i A P _ r a w ¯ M L i = M L _ r a w i M L _ r a w ¯ ( 4 ) \begin{equation} \label{eq:KINECAL_AP_ML} \begin{split} AP_i = AP\_raw_{i}-\bar{AP\_raw} \\ \\ ML_i = ML\_raw_{i}-\bar{ML\_raw} \end{split} \end{equation} (4)
R D i = ( M L i ) 2 + ( A P i ) 2 ( 5 ) \begin{equation} \label{eq:KINECAL_RD} RD_i = \sqrt{(ML_{i} )^2 + (AP_{i} )^2} \end{equation} (5)


Calculation of sway metrics

The following section gives details of the equations used to calculate the sway metrics. In these equations, _AP relates to the AP time series (AP). For the most part, these equations were also used to calculate the metrics in the ML and RD directions. The exceptions being MFREQ and the 95% confidence ellipse.

The most straightforward metric to understand, and to calculate, is the mean distance of the CoM. This is simply the mean absolute distance moved from the mean position of the CoM over the time of each trial. Equation (5) was used to calculate MDIST_AP

M D I S T _ A P = 1 n | A P i | ( 5 ) \begin{equation}\label{eq:MDIST_AP} MDIST\_AP = \frac{1}{n}\sum_{}^{} \left | AP_{i} \right | \end{equation} (5)

where AP is the AP time series, i is a single time step, and n is the total number of times steps in the time series. ML was used in place of AP to calculate the MDIST_ML. RD was used in place of AP to calculate the MDIST.

RMS Distance

The Root-Mean-Squared process removes the sign and gives more prominence to values further away from the mean CoM position. Equation (6) was used to calculate RDIST_AP.

R D I S T _ A P = 1 n A P i 2 ( 6 ) \begin{equation}\label{eq:RDIST_AP} RDIST\_AP = \sqrt{\frac{1}{n}\sum_{}^{}AP_{i}^{2}} \end{equation} (6)

Where AP is the AP time series, i is a single time step, and n is the total number of times steps in the time series. ML was used in place of AP to calculate the RDIST_ML. RD was used in place of AP to calculate the RDIST.

Total excursion (CoM Path Length)

The total excursion is calculated by summing the distance between successive time steps. This is also known as the CoM path length. Equation (7) was used to calculate TOTEX_AP.

T O T E X _ A P = | A P i + 1 A P i | ( 7 ) \begin{equation}\label{eq:TOTEX_AP} TOTEX\_AP = \sum_{}^{}\left |AP_{i+1} - AP_{i} \right | \end{equation} (7)

Where AP is the AP time series, and i is a single time step. ML was used in place of AP to calculate the TOTEX_ML. RD was used, in place of AP, to calculate the TOTEX.

Mean Velocity

Mean velocity is the total excursion divided by the time of the sample in seconds. Equation (8) was used to calculate the MVELO_AP.

M V E L O _ A P = T O T E X _ A P / t ( 8 ) \begin{equation}\label{eq:MVELO_AP} MVELO\_AP = TOTEX\_AP / t \end{equation} (8)

Where AP is the AP time series, and t is the time for the trial in seconds. ML was used in place of AP to calculate the MVELO_ML. RD was used in place of AP to calculate the MVELO.

Mean angular frequency AP and ML

MFREQ_AP is the frequency, in Hz, of a sinusoidal oscillation with an average value of the mean MDIST_AP and a total path length of TOTEX_AP. This was calculated using equation (9).

M F R E Q _ A P = M V E L O _ A P 4 2 M D I S T _ A P ( 9 ) \begin{equation}\label{eq:MFREQ_AP} MFREQ\_AP = \frac{MVELO\_AP}{4\sqrt{2MDIST\_AP}} \end{equation} (9)

Where MVELO_AP is defined by equation (8) and MDIST_AP is defined by equation (5). MVELO_ML was used in place of MVELO_AP, and MDIST_ML was used in place of MDIST_AP to calculate the MFREQ_ML.

Mean Angular Frequency

MFREQ is the mean angular frequency is the rotational frequency in Hz. This is the number of revolutions per second of the CoM if it had travelled the total excursion around a circle with a radius of the mean distance. It is calculated using values derived from RD, given by Equation (10).

M F R E Q = M V E L O 2 π M D I S T ( 10 ) \begin{equation}\label{eq:MFREQ} MFREQ = \frac{MVELO}{2\pi MDIST} \end{equation} (10)

Where MVELO is defined by equation (8) and MDIST is defined by equation (5).

95% confidence elliptical area

The AREA_CE is given by the set of equations below (11). The elliptical area is an estimate of the area described by the maximum and minimum AP and ML values of the time series. To reduce the effect of rapid changes in direction, this value is scaled to 1.96 standard deviations of the mean values.

p = c o v A P , M L σ A P σ M L A P r = 1 p σ A P 1.96 M L r = 1 + p σ M L 1.96 a r e a = π A P r M L r ( 11 ) \begin{equation}\label{eq:AREA_CE} \begin{aligned} p = \frac{cov_{AP, ML}}{\sigma_{AP} \sigma_{ML}} \\ \\ AP_r = \sqrt{1 - p} \sigma_{AP} 1.96 \\ \\ ML_r = \sqrt{1 + p} \sigma_{ML} 1.96 \\ \\ area = \pi AP_r ML_r \end{aligned} \end{equation} (11)

Where p is the Pearson correlation coefficient, MLr is the ML elliptical radius scaled to the 95% Confidence Interval (CI), and APr is the AP elliptical radius scaled to the 95% CI.


Data Description

KINECAL contains the recordings of 90 participants carrying out 11 movements, commonly used in the clinical assessment of balance impairment, frailty and falls-risk. Details of how each movement was carried out are shown in Table 1.

Demographics

register.csv provides the headline information about each participant along with demographic data and other labels. The file contains the following columns:

  • part_id: participant id
  • group: healthy-adult (HA), non-fallers (NF), single-faller, by falls history (FHs), multiple-faller, by falls history (FHm),
  • age: chronological age at the time of recording.
  • sex: biological sex (m/f)
  • height: height in meters
  • weight: weight in kg
  • BMI: Kg/m2
  • recorded_in_the_lab: indicates if the recording was made in the lab
  • clinically-at-Risk: indicates if the participants were marked as clinically at risk

Sway metrics

sway_metric.csv contains common sway metrics for each movement per participant. This file includes the following headers:

  • part_id: participant ID
  • movement: the movement that was recorded
  • group: non-faller, single-faller, by falls history (FHs), multiple-faller, by falls history (FHm),
  • age: Chronological age
  • sex: Biological sex
  • recorded_in_the_lab: indicates if the recording was made in the lab
  • clinically-at-Risk: Indicates if the participants were marked as clinically at risk
  • RDIST_ML: Value for RMS distance in ML direction
  • RDIST_AP: Value for RMS distance in AP direction
  • RDIST: Value for the overall RMS distance
  • MDIST_ML: Value for mean distance in ML direction
  • MDIST_AP: Value for man distance in AP direction
  • MDIST: Value for the overall mean distance
  • TOTEX_ML: Value for total exertion in ML direction
  • TOTEX_AP: Value for total exertion in AP direction
  • TOTEX: Value for the overall total exertion
  • MVELO_ML: Value for mean velocity in ML direction
  • MVELO_AP: Value for mean velocity in ML direction
  • MVELO: Value for the overall mean velocity 
  • MFREQ_ML: Value for mean frequency in ML direction
  • MFREQ_AP: Value for mean frequency in AP direction
  • MFREQ: Value for the overall mean frequency
  • AREA_CE: The area of the 95% confidence ellipsis

Depth data

A file has been created for each depth frame recorded by the Kinect camera (30 fps). The file stores the ushort values in binary format. The file is named DepthUshort<clock_tick>.bin. The binary file is a flattened 2D array of pixel by pixel distances from that camera to objects in the frame. This array can be resized back to the original 424, 512 aspect to reconstitute the original image. See the usage notes for python code that can achieve this.

Skeleton joint coordinates

In addition to raw depth data, the joint positions as defined by the Kinect 25 joint model has also been stored (30 fps). These are stored in plain text. The files are named <clock_tick>.txt. The file format is as follows:

  • Joint_name: The name for an individual joint, e.g. SpineBase
  • Tracked: This indicates if the joint was tracked or inferred.  Inferred positions happen in the case of occlusion
  • x-3D: The estimated x position in 3D space
  • y-3D: The estimated x position in 3D space
  • z-3D: The estimated x position in 3D space
  • x-pixel-pos: The x-pixel potion from the depth camera
  • y-pixel-pos: The y-pixel potion from the depth camera

Folder structure

This section provides details of the files and folders which make up the dataset. NB. The subfolders, shown for <part_id>_STS-5, also exist for the other movements. They are not repeated here, for brevity's sake.

kinecal
└── <part_id>
        ├──- <part_id>_STS-5
        │        ├── sway_metrics
        │        ├── skel
        │        │     ├── DepthUshort<clock_tick>.bin
        │        ├── depth
        │             └── <clock_tick>.txt
        ├── <part_id>_Quite-Standing-Eyes-Open
        ├── <part_id>_Quite-Standing-Eyes-Closed
        ├── <part_id>_Foam_Quite-Standing-Eyes-Open
        ├── <part_id>_Foam_Quite-Standing-Eyes-Closed
        ├── <part_id>_Semi-Tandem-Balance
        ├── <part_id>_Tandem-Balance
        ├── <part_id>_Unilateral-Stance-Eyes-Open
        ├── <part_id>_Unilateral-Stance-Eyes-Closed
        ├── <part_id>_3m-walk
        └── <part_id>_TUG

Usage Notes

Both raw depth and joint positions are provided. These files can be processed using many commonly used data processing tools, e.g. Python and Matlab.

The data can be used to study both postural sway metrics, akin to data derived from force plates and whole body metrics, such as joint angles, akin to metrics derived from marker-based motion capture, such as VICON. 

A sample subset of KIECAL is provided, along with Python code, demonstrating how to use KINECAL files (sample_set.zip). Note that the sample code requires the helper functions included in (sway_utils.zip), and these two archives should be unpacked into side-by-side folders. sway_utils can be used by any third party to speed up the process of using KINECAL data.  Run view_smple_files.py from sample_set.zip to visualise the sample data.

Although the largest of its kind, KINECAL would benefit from expansion. To this end, It is we will subsiquetly release the code used to capture the data. In this way, the project can naturally expand. We will be using this dataset in a set of upcoming papers. The first step to publishing this work is the publication of this dataset. As the papers are published, we will update this section of this project.


Release Notes

Version 1.0.3: A small subset of KINECAL has been included to help the adoption of the dataset.

Version 1.0.2: license amended.

Version 1.0.1: license amended.

Version 1.0.0: the initial release.


Ethics

The authors declare no ethics concerns. Ethical approval was obtained from the University Research Ethics Committee (ethics approval ref: 020517-ESS-CC(1), and ref: SE161757). All participants provided written informed consent.


Acknowledgements

Thank you to Manchester Metropolitan University for funding this research.


Conflicts of Interest

There are no conflict of interest.


References

  1. Daniel Leightley, Moi Hoon Yap, Jessica Coulson, Yoann Barnouin, Jamie S. McPhee, Y Barnouin J. Coulson, Jamie S. McPhee, Jessica Coulson, Yoann Barnouin, and Jamie S. McPhee. Benchmarking human motion analysis using kinect one: An open source dataset. In Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA), pages 1–7. IEEE, 12 2015
  2. Alexandre Bernardino, Christian Vismara, Sergi Bermudez i Badia, Elvio Gouveia, Fatima Baptista, Filomena Carnide, Simao Oom, and Hugo Gamboa. A dataset for the automatic assessment of functional senior fitness tests using kinect and physiological sensors. In 2016 1st In- ternational Conference on Technology and Innovation in Sports, Health and Wellbeing (TISHW), pages 1–6. IEEE, 12 2016.
  3. Marianna Capecci, Maria Gabriella Ceravolo, Francesco Ferracuti, Sabrina Iarlori, Andrea Monteriu, Luca Romeo, and Federica Verdini. The KIMORE Dataset: KInematic Assessment of MOvement and Clinical Scores for Remote Monitoring of Physical REhabilitation. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 27(7):1436– 1448, 7 2019.
  4. A. Shumway-Cook, S. Brauer, and M. Woollacott, “Predicting the Probability for Falls in Community-Dwelling Older Adults Using the Timed Up Go Test,” Phys. Ther., vol. 80, no. 9, pp. 896–903, Sep. 2000.
  5. Thomas E. Prieto, Joel B. Myklebust, Raymond G. Hoffmann, and Eric G. Lovett. Measures of postural steadiness: Differences between healthy young and elderly adults. IEEE Transactions on Biomedical Engineering, 43(9):956–966, 1996
  6. Ji-Won Kim, Gwang-Moon Eom, Chul-Seung Kim, Da-Hye Kim, Jae-Ho Lee, Byung Kyu Park, and Junghwa Hong. Sex differences in the postural sway characteristics of young and elderly subjects during quiet natural standing. Geriatrics & Gerontology International, 10(2):191– 198, 2 2010
  7. Jack M. Guralnik, Eleanor M. Simonsick, Luigi Ferrucci, Robert J Glynn, Lisa F Berkman, Dan G Blazer, Paul A Scherr, Robert B. Wallace, Guralnik, Ferrucci, Simonsick, Salive, and Wallace. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. Center for Successful Aging, 49(2):M85–M94, 3 1994
  8. ngo Michaelis, A. Kwiet, U. Gast, A. Boshof, T. Antvorskov, T. Jung, J. Rittweger, and D. Felsenberg. Decline of specific peak jumping power with age in master runners. Journal of musculoskeletal & neuronal interactions, 8(1):64–70, 2008
  9. Rachel E. Ward, Suzanne G. Leveille, Marla K. Beauchamp, Thomas Travison, Neil Alexander, Alan M. Jette, and Jonathan F. Bean. Func- tional performance as a predictor of injurious falls in older adults. Journal of the American Geriatrics Society, 63(2):315–320, 2 2015
  10. Diane Podsiadlo and Sandra Richardson. The Timed “Up & Go”: A Test of Basic Functional Mobility for Frail Elderly Persons. Journal of the American Geriatrics Society, 39(2):142–148, 2 1991

Share
Access

Access Policy:
Anyone can access the files, as long as they conform to the terms of the specified license.

License (for files):
Creative Commons Zero 1.0 Universal Public Domain Dedication

Corresponding Author
You must be logged in to view the contact information.
Versions

Files

Total uncompressed size: 180.1 GB.

Access the files
Folder Navigation: <base>/kinecal/80/80_STS-5
Name Size Modified
Parent Directory
depth
skel