Microsoft Kinect V2 Research Articles

AIR-Act2Act: Human–human interaction dataset for teaching non-verbal social behaviors to robots

To better interact with users, a social robot should understand the users’ behavior, infer the intention, and respond appropriately. Machine learning is one way of implementing robot intelligence. It provides the ability to automatically learn and improve from experience instead of explicitly telling the robot what to do. Social skills can also be learned through watching human–human interaction videos. However, human–human interaction datasets are relatively scarce to learn interactions that occur in various situations. Moreover, we aim to use service robots in the elderly care domain; however, there has been no interaction dataset collected for this domain. For this reason, we introduce a human–human interaction dataset for teaching non-verbal social behaviors to robots. It is the only interaction dataset that elderly people have participated in as performers. We recruited 100 elderly people and 2 college students to perform 10 interactions in an indoor environment. The entire dataset has 5,000 interaction samples, each of which contains depth maps, body indexes, and 3D skeletal data that are captured with three Microsoft Kinect v2 sensors. In addition, we provide the joint angles of a humanoid NAO robot which are converted from the human behavior that robots need to learn. The dataset and useful Python scripts are available for download at https://github.com/ai4r/AIR-Act2Act. It can be used to not only teach social skills to robots but also benchmark action recognition algorithms.

Determination of Body Parts in Holstein Friesian Cows Comparing Neural Networks and k Nearest Neighbour Classification.

Simple SummaryThe recognition of objects in three dimensional (3D) data is challenging, especially when it comes to the partition of objects into predefined segments. In this study, two machine learning approaches were applied to recognize the body parts head, rump, back, legs and udder of dairy cows in 3D data recorded with Microsoft Kinect V1, for usage in automated exterior evaluation. Five properties of data points were used as features in order to train a k nearest neighbour classifier and a neural network to determine the body part on which the point was located. Both algorithms used for this feature-based approach are computationally faster and more flexible compared to a model-based object detection priorly used for the same purpose. Both methods could be considered successful for the determination of body parts via pixel properties. Reaching high to very high overall accuracies and very small Hamming losses, the k nearest neighbour classification is superior to the neural network which reached medium to high evaluation metrics. However, k nearest neighbour classification is at runtime prone to higher costs regarding computational time and memory, while once trained, the neural network delivers the classification results very quickly. This needs also to be taken into account in the decision of which method should be implemented.Machine learning methods have become increasingly important in animal science, and the success of an automated application using machine learning often depends on the right choice of method for the respective problem and data set. The recognition of objects in 3D data is still a widely studied topic and especially challenging when it comes to the partition of objects into predefined segments. In this study, two machine learning approaches were utilized for the recognition of body parts of dairy cows from 3D point clouds, i.e., sets of data points in space. The low cost off-the-shelf depth sensor Microsoft Kinect V1 has been used in various studies related to dairy cows. The 3D data were gathered from a multi-Kinect recording unit which was designed to record Holstein Friesian cows from both sides in free walking from three different camera positions. For the determination of the body parts head, rump, back, legs and udder, five properties of the pixels in the depth maps (row index, column index, depth value, variance, mean curvature) were used as features in the training data set. For each camera positions, a k nearest neighbour classifier and a neural network were trained and compared afterwards. Both methods showed small Hamming losses (between 0.007 and 0.027 for k nearest neighbour (kNN) classification and between 0.045 and 0.079 for neural networks) and could be considered successful regarding the classification of pixel to body parts. However, the kNN classifier was superior, reaching overall accuracies 0.888 to 0.976 varying with the camera position. Precision and recall values associated with individual body parts ranged from 0.84 to 1 and from 0.83 to 1, respectively. Once trained, kNN classification is at runtime prone to higher costs in terms of computational time and memory compared to the neural networks. The cost vs. accuracy ratio for each methodology needs to be taken into account in the decision of which method should be implemented in the application.

Elderly Care Based on Hand Gestures Using Kinect Sensor

Technological advances have allowed hand gestures to become an important research field especially in applications such as health care and assisting applications for elderly people, providing a natural interaction with the assisting system through a camera by making specific gestures. In this study, we proposed three different scenarios using a Microsoft Kinect V2 depth sensor then evaluated the effectiveness of the outcomes. The first scenario used joint tracking combined with a depth threshold to enhance hand segmentation and efficiently recognise the number of fingers extended. The second scenario utilised the metadata parameters provided by the Kinect V2 depth sensor, which provided 11 parameters related to the tracked body and gave information about three gestures for each hand. The third scenario used a simple convolutional neural network with joint tracking by depth metadata to recognise and classify five hand gesture categories. In this study, deaf-mute elderly people performed five different hand gestures, each related to a specific request, such as needing water, meal, toilet, help and medicine. Next, the request was sent via the global system for mobile communication (GSM) as a text message to the care provider’s smartphone because the elderly subjects could not execute any activity independently.

Assessing the Performance of RGB-D Sensors for 3D Fruit Crop Canopy Characterization under Different Operating and Lighting Conditions.

The use of 3D sensors combined with appropriate data processing and analysis has provided tools to optimise agricultural management through the application of precision agriculture. The recent development of low-cost RGB-Depth cameras has presented an opportunity to introduce 3D sensors into the agricultural community. However, due to the sensitivity of these sensors to highly illuminated environments, it is necessary to know under which conditions RGB-D sensors are capable of operating. This work presents a methodology to evaluate the performance of RGB-D sensors under different lighting and distance conditions, considering both geometrical and spectral (colour and NIR) features. The methodology was applied to evaluate the performance of the Microsoft Kinect v2 sensor in an apple orchard. The results show that sensor resolution and precision decreased significantly under middle to high ambient illuminance (>2000 lx). However, this effect was minimised when measurements were conducted closer to the target. In contrast, illuminance levels below 50 lx affected the quality of colour data and may require the use of artificial lighting. The methodology was useful for characterizing sensor performance throughout the full range of ambient conditions in commercial orchards. Although Kinect v2 was originally developed for indoor conditions, it performed well under a range of outdoor conditions.

Tracking and Characterization of Spinal Cord-Injured Patients by Means of RGB-D Sensors.

In physical rehabilitation, motion capture solutions are well-known but not as widespread as they could be. The main limit to their diffusion is not related to cost or usability but to the fact that the data generated when tracking a person must be elaborated according to the specific context and aim. This paper proposes a solution including customized motion capture and data elaboration with the aim of supporting medical personnel in the assessment of spinal cord-injured (SCI) patients using a wheelchair. The configuration of the full-body motion capturing system is based on an asymmetric 3 Microsoft Kinect v2 sensor layout that provides a path of up to 6 m, which is required to properly track the wheelchair. Data elaboration is focused on the automatic recognition of the pushing cycles and on plotting any kinematic parameter that may be interesting in the assessment. Five movements have been considered to evaluate the wheelchair propulsion: the humeral elevation, the horizontal abduction of the humerus, the humeral rotation, the elbow flexion and the trunk extension along the sagittal plane. More than 60 volunteers with a spinal cord injury were enrolled for testing the solution. To evaluate the reliability of the data computed with SCI APPlication (APP) for the pushing cycle analysis, the patients were subdivided in four groups according to the level of the spinal cord injury (i.e., high paraplegia, low paraplegia, C7 tetraplegia and C6 tetraplegia). For each group, the average value and the standard deviation were computed and a comparison with similar acquisitions performed with a high-end solution is shown. The measurements computed by the SCI-APP show a good reliability for analyzing the movements of SCI patients’ propulsion wheelchair.

Collision Avoidance Interaction Between Human and a Hidden Robot Based on Kinect and Robot Data Fusion

Human-Robot Interaction (HRI) is a largely addressed subject today. In order to ensure co-existence and space sharing between human and robot, collision avoidance is one of the main strategies for interaction between them without contact. It is thus usual to use a 3D depth camera sensor (Microsoft Kinect V2) which may involve issues related to occluded robot in the camera view. While several works overcame this issue by applying infinite depth principle or increasing the number of cameras, in the current work we developed and applied an original new approach that combines data of one 3D depth sensor (Kinect) and proprioceptive robot sensors. This method uses the principle of limited safety contour around the obstacle to dynamically estimate the robot-obstacle distance, and then generate the repulsive force that controls the robot. For validation, our approach is applied in real time to avoid collisions between dynamical obstacles (humans or objects) and the end-effector of a real 7-dof Kuka LBR iiwa collaborative robot. Our method is experimentally compared with existing methods based on infinite depth principle when the robot is hidden by the obstacle with respect to the camera view. Results showed smoother behavior and more stability of the robot using our method. Extensive experiments of our method, using several strategies based on distancing and its combination with dodging were done. Results have shown a reactive and efficient collision avoidance, by ensuring a minimum obstacle-robot distance (of $\approx \text{240 mm}$ ), even when the robot is in an occluded zone in the Kinect camera view.

Accurate body measurement of live cattle using three depth cameras and non-rigid 3-D shape recovery

Body condition scoring of livestock is widely used as a subjective method for assessing energy reserves and making management decisions for livestock. Since animal dimensions are often measured manually, the procedure is time-consuming, expensive and stressful for both the farmer and animal. Recent advances in three-dimensional sensor technology provide innovative tools for the design of automated contactless systems for assessing the animal body condition. The objective of this paper is to design an automated computer vision system capable to generate an accurate three-dimensional model of live cattle. The system is based on a non-rigid 3-D shape reconstruction utilizing data from depth cameras. The design methodology includes three Microsoft Kinect v2 cameras, computer vision, signal filtering of point clouds, pattern recognition using 3-D feature extraction techniques, and statistical analysis using point and interval estimations. The quality of generated three-dimensional body models is validated against manually measured nine references, such as withers height, hip height, chest depth, oblique body length, heart girth, etc. With a 90% confidence level, measurement errors in the proposed system among all measured estimates are less than 3%. Experimental results show that the proposed approach can serve as a new accurate method for non-contact body measurement of livestock.

Three-dimensional measurements of tree crown movement using an infrared time-of-flight camera

The study of wind–tree interaction represents a challenge in fluid dynamics partly because the three-dimensional tree crown and its branch structure are dynamically adapting to the mean and turbulent wind components. With the advent of infrared technologies these complex, three-dimensional, real time interactions between wind and the response of a tree can be assessed. The present work makes use of a Microsoft Kinect V2, which incorporates an infrared time-of-flight camera, to capture images containing point cloud data of a single garden tree in turbulent winds. The raw output from the sensor is used to translate three-dimensional information into mean crown positions and projected frontal canopy areas. This novel infrared time-of-flight sensor technique facilitates a better understanding of the behaviour of the tree crown against wind speed. Overall, less movement and larger crown areas are observed at lower wind speeds. As the wind speed increases, a general shrinkage in area, more swaying and a smoothing of the crown edges are noted. Correlations between the observed wind force exerted on the tree and the displacement of the crown centre as well as between the wind force and the projected frontal canopy area are determined. The displacement and the drag force are positively correlated for all wind speeds. The drag force and frontal canopy area are positively correlated at low wind speeds and negatively correlated at higher wind speeds with a bi-stable phase between. The infrared time-of-flight sensor has demonstrated its capacity to provide real-time three-dimensional scanning of trees in a wind testing facility providing new insights into the complex wind–tree interaction problem.

Guest Editors’ Introduction to the Special Issue on RGB-D Vision: Methods and Applications

The twenty-six papers in this special issue focus on Red Blue Green (RBG)-D vision, an emerging research topic in computer vision, with a number of applications in robotics, entertainment, biometrics and multimedia. Compared to 2D images and 3D data (including depth images, point clouds and meshes), RGB-D images represent both the photometric and geometric information of a scene. Moreover, low-cost consumer depth cameras (e.g., Microsoft Kinect v2, Intel Realsense, Orbbec Astra) can enable realtime applications due to their high acquisition frame-rate. In the last few years, a large number of RGB-D datasets have also been publicly released to tackle various vision tasks. Although remarkable progress has been achieved, several critical problems still remain open. The aim of this special issue is to stimulate researchers from different fields to present their state-of-the-art work, and to provide a cross-fertilization ground for discussions on the next steps in this important research area.

Automated classification of movement quality using the Microsoft Kinect V2 sensor

Practitioners commonly perform movement quality assessment through qualitative assessment protocols, which can be time-intensive and prone to inter-rater measurement bias. The advent of portable and inexpensive marker-less motion capture systems can improve assessment through objective joint kinematic analysis. The current study aimed to evaluate various machine learning models that used kinematic features from Kinect position data to classify a performer's Movement Competency Screen (MCS) score. A Kinect V2 sensor collected position data from 31 physically active males as they performed bilateral squat, forward lunge, and single-leg squat; and the movement quality was rated according to the MCS criteria. Features were extracted and selected from domain knowledge-based kinematic variables as model input. Multiclass logistic regression (MLR) was then performed to translate joint kinematics into MCS score. Performance indicators were calculated after a 10-fold cross validation of each model developed from Kinect-based kinematic variables. The analyses revealed that the models' sensitivity, specificity, and accuracy ranged from 0.66 to 0.89, 0.58 to 0.86, and 0.74 to 0.85, respectively. In conclusion, the Kinect-based automated movement quality assessment is a suitable, novel, and practical approach to movement quality assessment.

The Accuracy of the Microsoft Kinect V2 Sensor for Human Gait Analysis. A Different Approach for Comparison with the Ground Truth.

Several studies have examined the accuracy of the Kinect V2 sensor during gait analysis. Usually the data retrieved by the Kinect V2 sensor are compared with the ground truth of certified systems using a Euclidean comparison. Due to the Kinect V2 sensor latency, the application of a uniform temporal alignment is not adequate to compare the signals. On that basis, the purpose of this study was to explore the abilities of the dynamic time warping (DTW) algorithm to compensate for sensor latency (3 samples or 90 ms) and develop a proper accuracy estimation. During the experimental stage, six iterations were performed using the a dual Kinect V2 system. The walking tests were developed at a self-selected speed. The sensor accuracy for Euclidean matching was consistent with that reported in previous studies. After latency compensation, the sensor accuracy demonstrated considerably lower error rates for all joints. This demonstrated that the accuracy was underestimated due to the use of inappropriate comparison techniques. On the contrary, DTW is a potential method that compensates for the sensor latency, and works sufficiently in comparison with certified systems.

Kinect v2 tracked Body Joint Smoothing for Kinematic Analysis in Musculoskeletal Disorders.

Body joint monitoring is essential for disorder diagnosis and assessment of treatment effectiveness. Microsoft Kinect v2 is a low-cost and markerless human motion-tracking RGB-D sensor that provides spatial locations of tracked skeletal joints in the form of 3D coordinates. Sometimes, body tracking of kinect v2 produces erratic 3D coordinates, which affects the real-time tracking performance of the sensor. A careful study of the literature suggests that skeletal tracking of kinect v2 needs further exploration. This work proposes a filter combined with the concept of body kinematics to remove noise and enhances the quality of 3D coordinates in body frame data. Also, it generates "Motion Signature" of the tracked joint, which shows movement pattern for kinematic analysis, and helpful in joint monitoring of Musculoskeletal Disorders (MSD). The clinically relevant anatomical movement was executed, to evaluate the performance of the proposed filter. We compared Range of Motion (RoM) values obtained from the proposed filter with the gold standard goniometry. Results indicate that RoM values from the proposed filter are in high correlation with the goniometry with an Intraclass Correlation Coefficient values ranging between 0.95 to 0.98 authenticating that it improves the skeletal joint tracking of kinect v2.

Automatic measurement of the hand dimensions using consumer 3D cameras

<p class="Abstract">This article describes the metrological characterisation of two prototypes that use the point clouds acquired by consumer 3D cameras for the measurement of the human hand geometrical parameters. The initial part of the work is focused on the general description of algorithms that allow for the derivation of dimensional parameters of the hand. Algorithms were tested on data acquired using Microsoft Kinect v2 and Intel RealSense D400 series sensors. The accuracy of the proposed measurement methods has been evaluated in different tests aiming to identify bias errors deriving from point-cloud inaccuracy and at the identification of the effect of the hand pressure and the wrist flexion/extension. Results evidenced an accuracy better than 1 mm in the identification of the hand’s linear dimension and better than 20 cm<sup>3 </sup>for hand volume measurements. The relative uncertainty of linear dimensions, areas, and volumes was in the range of 1-10 %. Measurements performed with the Intel RealSense D400 were, on average, more repeatable than those performed with Microsoft Kinect. The uncertainty values limit the use of these devices to applications where the requested accuracy is larger than 5 % (volume measurements), 3 % (area measurements), and 1 mm (hands’ linear dimensions and thickness).</p>

Applicability of a Single Depth Sensor in Real-Time 3D Clothes Simulation: Augmented Reality Virtual Dressing Room Using Kinect Sensor

A busy lifestyle led people to buy readymade clothes from retail stores with or without fit-on, expecting a perfect match. The existing online cloth shopping systems are capable of providing only 2D images of the clothes, which does not lead to a perfect match for the individual user. To overcome this problem, the apparel industry conducts many studies to reduce the time gap between cloth selection and final purchase by introducing “virtual dressing rooms.” This paper discusses the design and implementation of augmented reality “virtual dressing room” for real-time simulation of 3D clothes. The system is developed using a single Microsoft Kinect V2 sensor as the depth sensor, to obtain user body parameter measurements, including 3D measurements such as the circumferences of chest, waist, hip, thigh, and knee to develop a unique model for each user. The size category of the clothes is chosen based on the measurements of each customer. The Unity3D game engine was incorporated for overlaying 3D clothes virtually on the user in real time. The system is also equipped with gender identification and gesture controllers to select the cloth. The developed application successfully augmented the selected dress model with physics motions according to the physical movements made by the user, which provides a realistic fitting experience. The performance evaluation reveals that a single depth sensor can be applied in the real-time simulation of 3D cloth with less than 10% of the average measurement error.

The reliability and reproducibility of sagittal spinal curvature measurement using the Microsoft Kinect V2.

Abnormal sagittal spinal curvature is associated with pain, decreased mobility, respiratory problems and increased mortality. Time-of-flight technology of the Microsoft Kinect sensor can reconstruct a three-dimensional image of the back quickly and inexpensively. To estimate the extent of the reproducibility of sagittal spine curvature measurement using the Microsoft Kinect sensor. Simultaneous measurement of thoracic and lumbar spine using the Microsoft Kinect sensor in 37 participants. Two investigators gave standardised instructions and each captured 3 images. Thoracic kyphosis and lumbar lordosis angle indexes were calculated using maximum height divided by the length. Adult participants (mean age in years (SD) = 51.7 (20.6); 57% female; BMI in kg/m2 (SD) = 24.9 (3.3)) kyphosis and lordosis indexes showed high intra-rater and inter-rater ICC values (0.960-0.973). The means of the first images from both raters had significantly larger kyphosis indexes compared to the second and third images, yet no difference between means in lordosis data. The results indicate that the Microsoft Kinect sensor has a reproducible method with high intra-rater and inter-rater reliability. The difference between the means over repeated measures suggest the second image capture is more consistent. It is a reproducible and quick method in clinical and research settings.

Accuracy evaluation of two markerless motion capture systems for measurement of upper extremities: Kinect V2 and Captiv

AbstractMotion capturing is a promising method to assess working postures and human movements and, therewith, the risk of musculoskeletal injuries that could occur while performing manual tasks in industrial settings. To obtain a reliable risk assessment, the motion capture system used has to accurately measure body postures adopted by the worker during the task. This study evaluates the accuracy to measure joint angles of upper extremities of two different motion capture systems, namely the Microsoft Kinect V2 and the Captiv system, for angles of upper extremities. For this purpose, an experimental study was conducted involving 12 subjects performing preset static postures and basic movements, including elbow flexion, shoulder flexion, and shoulder abduction. In addition, to examine whether self‐occlusion or occlusion of body parts by work equipment has an impact on the accuracy of the Kinect V2, the subjects handled boxes during some of the tests. As a gold standard, a goniometer for static and an angle scale for dynamic exercises was used. The Captiv system shows high correlation coefficients (r > .93) and small mean absolute errors (<5°) for all movements except for elbow flexion. The Kinect V2 has sufficient results for joint angles captured without occlusion as well, but the accuracy significantly decreases when occlusion occurs.

Accuracy of image data stream of a markerless motion capture system in determining the local dynamic stability and joint kinematics of human gait

Assessment of gait parameters is commonly performed through the high-end motion tracking systems, which limits the measurement to sophisticated laboratory settings due to its excessive cost. Recently, Microsoft Kinect (v2) sensor has become popular in clinical gait analysis due to its low-cost. But, determining the accuracy of its RGB-D image data stream in measuring the joint kinematics and local dynamic stability remains an unsolved problem. This study examined the suitability of Kinect(v2) RGB-D image data stream in assessing those gait parameters. Fifteen healthy participants walked on a treadmill during which lower body kinematics were measured by a Kinect(v2) sensor and a optophotogrametric tracking system, simultaneously. Extended Kalman filter was used to extract the lower extremity joint angles from Kinect, while inverse kinematics was used for the gold standard system. For both systems, local dynamic stability was assessed using maximal Lyapunov exponent. Sprague’s validation metrics, root mean square error (RMSE) and normalized RMSE were computed to confirm the difference between the joint angles time series of the two systems while relative agreement between them was investigated through Pearson’s correlation coefficient (pr). Fisher’s Exact Test was performed on maximal Lyapunov exponent to investigate the data independence while reliability was assessed using intraclass correlation coefficients. This study concludes that the RGB-D data stream of Kinect sensor is efficient in estimating joint kinematics, but not suitable for measuring the local dynamic stability.

AUTSL: A Large Scale Multi-Modal Turkish Sign Language Dataset and Baseline Methods

Sign language recognition is a challenging problem where signs are identified by simultaneous local and global articulations of multiple sources, i.e. hand shape and orientation, hand movements, body posture, and facial expressions. Solving this problem computationally for a large vocabulary of signs in real life settings is still a challenge, even with the state-of-the-art models. In this study, we present a new largescale multi-modal Turkish Sign Language dataset (AUTSL) with a benchmark and provide baseline models for performance evaluations. Our dataset consists of 226 signs performed by 43 different signers and 38,336 isolated sign video samples in total. Samples contain a wide variety of backgrounds recorded in indoor and outdoor environments. Moreover, spatial positions and the postures of signers also vary in the recordings. Each sample is recorded with Microsoft Kinect v2 and contains RGB, depth, and skeleton modalities. We prepared benchmark training and test sets for user independent assessments of the models. We trained several deep learning based models and provide empirical evaluations using the benchmark; we used CNNs to extract features, unidirectional and bidirectional LSTM models to characterize temporal information. We also incorporated feature pooling modules and temporal attention to our models to improve the performances. We evaluated our baseline models on AUTSL and Montalbano datasets. Our models achieved competitive results with the state-of-the-art methods on Montalbano dataset, i.e. 96.11% accuracy. In AUTSL random train-test splits, our models performed up to 95.95% accuracy. In the proposed user-independent benchmark dataset our best baseline model achieved 62.02% accuracy. The gaps in the performances of the same baseline models show the challenges inherent in our benchmark dataset. AUTSL benchmark dataset is publicly available at https://cvml.ankara.edu.tr.

Tai Chi Care: An Exergaming Software using Microsoft Kinect V2 for Blind or Low Vision Person during Confinement

Blind or low vision people need to practice activities for their mental and physical health to minimize the risk of suffering from articulation pain but they have problems due to difficulties and inaccessibility of displacement especially during the COVID-19 pandemic where everyone in this world was asked to stay at home during confinement. To solve these problems, we have developed a software tool for a care Tai Chi exergaming to encourage them to practice exercise at home using body tracking by Microsoft Kinect V2 and audio feedback. This software acts as a Tai Chi treatment, teaches four poses, and has a customized audio feedback to help person to understand each pose and generates progress graphs to evaluate the success of these exercises. We used the SDK libraries of the Kinect to obtain 3D joint position from sensors of the Kinect to calculate the angles and distances between joints to help the person to position in front of the Kinect, evaluate the different gestures of flexions and extensions of knees and elbows of each exercises, and body balance direction to avoid falling risk. These exercises have been evaluated with persons who are blind or low vision to improve feasibility and feedback.

Improvement of Body Movements and Stability of Blind or Visually Impaired Adults by Physical Activity using Kinect V2

People who are blind or low vision need to follow activities routines for their mental and physical health to minimize the risk of suffering from bleeding in articulation but they have problems due to difficulties and inaccessibility of displacement. This paper introduces and evaluate a set of exercises to improve the bodily movement and stability using body tracking by Microsoft Kinect V2 and audio feedback. These exercises are composed of a sequence of different postures, has an audio feedback personalized to help people to understand each gestures and can correct them if it is not correct, and generates a summary graph to evaluate the success rate of exercises. To obtain the 3D joint coordinates from the depth sensor, we used the SDK V2.0 of the Microsoft Kinect. We use these coordinates to calculate the distances and angles between joints of interest firstly to position the user in the area field of the Kinect sensor, evaluate the different postures of movements of knees, elbows and shoulders, and detect the body balance if he is leaning and in which direction to avoid falling. These physical exercises have been evaluated to improve feasibility and feedback with persons who are blind or low vision.