An Automated Child Gait Monitoring System
Uthpala Samarakoon E.M.H.J.Ekanayake
Senior Lecture: Departmant of Information Technology Student: Department of Information Technology
Sri Lanka Institute of Information Technology (SLIIT) Sri Lanka Institute of Information Technology (SLIIT)
Malabe, Sri Lanka Malabe, Sri Lanka
[email protected] [email protected]

A.M.U.E.Attanayake A.I.Sellahannadi
Student: Department of Information Technology Student: Department of Information Technology
Sri Lanka Institute of Information Technology (SLIIT) Sri Lanka Institute of Information Technology (SLIIT)
Malabe, Sri Lanka Malabe, Sri Lanka
[email protected] [email protected]

D.J chinnah
Student: Department of Information Technology
Sri Lanka Institute of Information Technology (SLIIT)
Malabe, Sri Lanka
[email protected]

Abstract – The Walk is the most important part of the neurological examination. Gait refers to the style of walking of a person. Then there are many gait abnormalities. A child normally begins to walk between 12 and 14 months, then very important to identify if there is an abnormality in child walking pattern. Observation of these abnormalities is an important aspect of diagnosis they may provide information about several musculoskeletal and neurological conditions. For the diagnosis purpose doctor needs to go through with some parameters manually, like stride length, step distance and space between two knees then they come up with the suggestion if there is an abnormality or normal. If details are abnormal doctor should decide what type of gait abnormality that child has, then the doctor should observer child walking pattern very carefully. Based on eye measurements and their experience they come up with the conclusion. Since our main focus is on recognizing above parameters and analysis the walking pattern automatically the research is gathered information, analyzed, developed and implemented. By our system it is possible do automatically. It will provide distance between two knees automatically, Step distance, step count, angle of the foot placement and system monitoring child’s walking pattern, then based on that parameters system will identify what type of gait abnormality that child has occurred and what is the disease the gait abnormality belongs to. Several techniques and methodology are used to interconnect with C# such as OpenCV, Kinect SDK in order to achieve the objectives of the application. Kinect Device is used to capture the body moments and gestures.
Key words -,OpenCV, Kinect software development kit, Gait, gait parameters


A walk is a human need and wants that an individual should have the ability to go wherever he or she needs to finish things except if the place he or she needs to restrain. Walk unpredictable people stand up to different issues while getting to a place elsewhere. It intensely effects to make their day to day work. Slipping is the genuine purpose behind falling for the energetic or more prepared individuals in the midst of strolling and besides sooner or later they can be completely serious back torment in the midst of the strolling. It is basic to clear up the purpose behind slipping and the agony. At some point, because of strange strolling designs they might be slipping or can have torment.
Gait is the manner by which a man stroll. Gait has two phases swing phase and stance phase. In the swing, phase happens when the foot isn’t in contact with the ground and stance phase happens when the foot contact with the ground. It is distinctive between normal gait pattern and the abnormal gait pattern.
There are many gait analysis frameworks which can be utilized to recognize the walking pattern of a person. Regardless, a large segment of them are utilized for security reason and some are exceptionally costly. Frameworks utilized for security reason can’t be utilized for a medicinal reason since specialists require nitty-gritty data as for the progression designs and the parameters.
For example, in a security reason stride examination framework simply perceived the human strolling design, it isn’t given any positive information as the yield. Regardless, for affliction conclusion reason authority requires bits of knowledge about some crucial walk factors like foot count, stride length angle of foot placement.
As indicated above, existing frameworks have confinements therefor proposed framework address above issues, restrict the obstacle and issues has a better than average enthusiasm for the framework.
In this proposed an automated child gaits monitoring system, the researchers are centred around tending to those issues and limitation along with the accompanying highlights; non-intrusive, low cost, movable system that can accurate measure a wide range of gait parameters using Kinect sensor with Software Development kit(SDK). Identifying abnormalities of the gait patterns, calculating stride length, distance between two knees, Step distance, step count and the angle of the foot placement, then based on those parameters system will distinguish what kind of gait irregularity that child has happened and particular malady the gait variation from the norm belongs to. By then it is fundamental for the star to close the issue as opposed to irritating steps, assessing separation.
In Sri Lanka increment number of individuals with walking abnormalities out of 100 people, 39 people ae with walking disabilities 1. By considering above statistics, a conclusion has been housekeeper that impressive measurer of the population in Sri Lanka is having gait abnormalities.
The proposed gaits monitoring system is target for children between age five to twelve. Because until a child is approximately three years old, their normal gait does not resemble that of an adult. Initially, there is a wide-based stance with rapid cadence and short steps, therefore our target age range is year five to year twelve 2.
There are many gait abnormalities but some are not affected by children, we selected frequently apparent gait abnormalities in the children.
Selected gait abnormalities
• Tip toe gait
• Waddling gait
• Hemiplegic gait
• Parkinsonian gait


Though there are so many gait recognition systems, they still lack useful features which are necessary for people. The proposed system will be a fully functioned system which has features of the existing system as sub-functions. The research group assured that the new system will be more beneficial for medical treatments. The table below shows a comparison between the existing system and the proposed system. (See figure 1)

Figure 1: Product Comparison

A) High Level Architectures and Implementation
We propose a motion vision approach for our gait recognition system. In motion vision there are two techniques motion base technique and model base technique. The proposed system used above two techniques 3.
Motion base technique
Motion base technique is a compact representation of to characterize the motion pattern of the human body without taking into consideration of model structure 3.Using motion based technique can analyse the four gait abnormalities.
• Analyse Waddling gait: – Should identify the pattern of Waddling (Myopathic) gait (patterns walk like a duck). Measure distance by foot, identify the angle of the knee sway and capture the gait cycle. Movements of the gait cycle compare with waddling gait movements then identified the abnormalities belongs to waddling gait or not.

• Analyse tip toe gait: – Should identify the pattern of tip toe gait. Identify there is no contact with a heel to floor during the gait cycle and capture the gait cycle. Movements of the gait cycle compare with tip toe gait movements then identified the abnormalities belongs to tip toe gait or not.

• Analyse hemiplegic gait: – Should identify the pattern of hemiplegic gait. Identify step is rotated away from the body, then towards it, forming a semicircle (limb moves in a semicircle with each step) and capture the gait cycle. Movements of the gait cycle compare with hemiplegic gait movements then identified the abnormalities belongs to hemiplegic gait or not.
• Analyse parkinsonian gait: – Should identify the pattern of parkinsonian gait. Measure distance between two steps, identify arm swing, count footsteps and capture the gait cycle. Movements of the gait cycle compare with parkinsonian gait movements then identified the abnormalities belongs to parkinsonian gait or not.

Model base technique
The model base technique is model matching in each walking sequence frame so that parameter such as trajectories are measured according to the model. It is studied static and dynamic body parameters and it will model the human skeleton. Model based approach is usually viewpoint and scale invariant 4. Then the model is appropriate to every frame of walking sequence and its gait parameters such as stride length, angle of the foot placement, distance between two steps, dynamic base. For capture the skeleton we used Kinect Sensor for XBOX 360 model 1414 with the Kinect SDK v1.0 beta 2.

Kinect XBOX 360 model 1414

Kinect is a camera-based, motion sensing input device developed by Microsoft. It provides full –body 3D motion capture face tracking, provides a high quality skeletal model in front of the Kinect sensor 5 in a Carte-Sian coordinate system, skeleton tracking and 20 skeletal joints from head over hips to the feet. (Hip, knee, foot, and head). (See Figure 2 and 3). Kinect provides approximately 30 skeleton frames per second.

Figure 2:

Figure 3: Kinect Xbox 360

Figure 4-Kinect skeleton joints

In our proposed system, there is three major components capture the skeleton and joint information using Kinect, extract the features from the skeleton information and finally identify the abnormality basis on previously recorded data using machine learning. (See Figure 4 and 5).

Figure 5: Overall Structure of the Proposed System

Figure 6: Flow Control Diagram

? capture the skeleton

We connected the Kinect sensor to a single PC running Windows 8.1 connected via USB 2.0.The sensor was placed at an approximate 45 degree angle and the height of 30 cm – 60cm above the floor. Then the person walked in front of the Kinect sensor. Sensor will recording the walking pattern and it will extracted the real time skeleton data via Microsoft Kinect SDK v1.8 (measurements of the joints and their angles) and save color or depth images of walking pattern sequence.

? Feature extraction

Gait is sustain periodic process, therefore need to analyzed gait cycle and gait parameters. We consider skeletal joints head to foot. Because Hemiplegic gait and Parkinsonian gait will depend on the full body and the Waddling gait and tip toe gait will depend on the lower part of the body Using feature extraction we extracted ten following gait parameters based on the Kinect provided 3D joints of the skeleton. (See figure 3)

? Left hip angle
? Right hip angle
? Left knee angle
? Right knee angle
? Left ankle angle
? Right ankle angle
? Distance between two knees
? Angle between two feet

To extract above mentioned gait parameters we need to extract some special type of features such as Static, Dynamic, Angular features 6.

• Static feature – Static features in terms of distance between joint-points 6.(See Table 1)

Left lower Leg distance (KNEE_LEFT, ANKLE_LEFT)
Right upper arm distance (ELBOW_RIGHT, SHOULDER_RIGHT)
Distance between two knees distance (KNEE_LEFT,KNEE_RIGHT)

Table 1: Static feature
• Angular feature – means while walking different parts of the body (side-view) make distinctive angles. Some of the angle in above mentioned features (Right hip angle, Left knee angle) are estimate based on the 3D coordinates of the three related joints provide by Kinect sensor. And some of the features (Angle between two feet) are estimated using two planes are first computed with three joints and the angle of interests is determine as the angle between the two normal vectors of two panes (See figure 6)

Figure 7: Angle feature

• Dynamic Features- Speed and the stride length of the human from the dynamic feature. Stride length is median of step length. The distance (in X-coordinate) between the alternate maxima’s in this plot gives the step lengths 6. (see figure 7) The number of frames in a stride and using the Kinect’s frame rate as 30 fps.

Figure 8: An example of half-gait-cycle extraction.

Gait cycle

Strat from initial contact of one foot with the ground to the subsequent contact of the same foot. Each stride (gait cycle) is composed of a stance phase where the foot is on the ground, followed by a swing phase where the foot is swing forward. Half-Gait-Cycle Detection Consider the plot of the absolute difference of X-coordinates between left and right ankle joint-points over consecutive frames. The half-gait-cycle is defined as the frames between two consecutive local minima in this plot 6. (See figure 7).

B) Technologies used:
OpenCV, Microsoft Visual studio 2015, Kinect Software Development Kit version1.0 beta 2,C

The paper proposed a fully automated child gait monitoring system for age between year five to year twelve. Clinical instructors, doctors can easily use our system with having less knowledge about information technology. Our main goal is to identify four gait abnormalities such as Waddling gait, Parkinsonian gait, Hemiplegic gait, Tip toe gait and gait parameters. The system will identify what type of gait abnormality that child has occurred and respective disease the gait abnormality belongs to. By then it is basic for the pro to close the issue instead of upsetting steps, evaluating detachment. It is smarter to recognize the irregularities in their childhood, at that point doctors can treat as indicated by the variations from the norm.
The system is fully automated then it is easy for the doctor to diagnose the problem rather than bothering with footsteps, measuring distance manually using a ruler and observe walking pattern normally. And also if the parent uses this system they can get a brief idea of what kind of disease their children suffer. It is anticipated that the collection of data over a long period of time will produce important facts concerning the development of gait patterns in children.
There is some existing system to analyses the walking pattern of a human which is highly expensive and most uses for the security purpose. Our system target for medical uses. We believe that our results from gait recognition with Kinect are very promising and highly reliable for them.
Our future work to identify all these kind gaits types and avoid them. Also keen to maximize age group and also try to use our technique to other field as well. Example like sports field. It could be consisting of create a user-friendly software for runners to provide them with real-time feedback regarding their running performance.


First of all we would like to thank our university Sri Lanka Institute of Information Technology (SLIIT) for giving this opportunity to develop and expose our skills and abilities via performing a research. This module is the best platform to show our talent to the world. We would like to sincerely thank our supervisor Mrs. Uthpala Samrakoon for the valuable insights and supported us to make this a success. This project would not have been possible without the guidance of our supervisor. Her guidance and concern provided throughout the research of our project is thoroughly appreciated in proper manner. Finally we would like to express our thanks to the people who support numerous ways in successful completion of this project within the limited time.


1 Mandulee Mendis,”Increasing number of Cerebral Palsy patients in Sri Lanka”, October10,2014.Online,Available: Accessed March 3, 2018.
2 Mary Lowth,” Gait Abnormalities in Children,” 12 Aug 2014,Available:

3 C.-C. Yang and Y.-L. Hsu. A review of accelerometry-based wearable motion detectors for physical activity monitoring. Sensors,10(8):7772–7788, 2010.

4 J. Wang, M. She, S. Nahavandi, and A. Kouzani. AReview of Vision-Based Gait Recognition Methods for Human Identi_cation. 2010 International Conference on Digital Image Computing: Techniques and
Applications, pages 320{327, Dec. 2010.
5 app Available at: 10 July. 2018

6 Mallick, Tanwi, Ankit Khedia, Partha Pratim Das, and Arun Kumar Majumdar. “Fast Gait Recognition from Kinect Skeletons.” In VISIGRAPP (3: VISAPP), pp. 342-349. 2016.