Non-visual Motion Tracking

Non-visual Motion TrackingAddition each(prenominal)y, (Taylor et al. 2010) demonstrated that the OSSCA method, which employs a combined intent of OCST, SCoRE, and SARA techniques to offshoot marker information and allows the estimation of roast parameters from kinematic information alone, without the exigency to use generic anatomical relationship assumptions, returns more reliable, repeatable and reproducible results than a standard generic regression approach.Although the accuracy of the data acquired by instrument of optical motion capture outlines is very high in the controlled surround of the lab, the ambulatory use of this fictitious character of equipment is cumbersome and presents signifi stinkpott limitations which stand not only compromise the precision of the acquired data, e.g. dependency on line-of-sight, limited orbit and latency of data (Schepers et al. 2010), but also the practicability of the acquisition itself, e.g. necessity of power source, good deal -up time, outdoor calibration of the dust.Non-visual motion introduce Non-visual motion tracking is a sensor establish technique, which pot be carried out, amongst others, with acoustic, magnetized, or inertial sensors, or with a combination of these methods.Ultrasound based acoustic organisations, e.g. the quiver system (Ward et al. 1997), Vallidis (Hazas and Ward 2002), the Cricket location system (Priyantha et al. 2000) and WearTrack (Foxlin and Harrington 2000), are loose of tracking the locations of pulse emitting beckons by victimization the time-of-flight breeding of audio signals.This type of motion tracking system is receiving set, however, as with visual motion tracking, settlement of the signal emitter poses a signifi tummyt limitation.In contrast, magnetized systems, e.g. MotionStar (Ascension Technology), are undetermined of estimating their amaze and orientation course within the global mastermind system, by use information from the local magnetic envi ronment, and are, thitherfore, not constricted by line-of-sight. However, these systems are very medium to ferromagnetic interferences.Inertial motion capture systems, e.g. Moven (Xsens Technologies) and dashing (Verhaert), employ the use of accelerometer and gyros to measure inclination travels. These systems are highly accurate, however, sensitive to vibration and flying field to integration drift over time.In fact, passim the past decade, the use of inertial sensors has gained increased popularity within researchers (Foxlin 1996 Roetenberg et al. 2007a Roetenberg et al. 2005 Roetenberg et al. 2009 Roetenberg et al. 2003 Roetenberg et al. 2007b Roetenberg and Veltink 2005), as closely as general population.Many people schedule their daily exercise based on the data presented by certain applications on their smartphones (e.g. wellness app, Argus, MyFitnessPal), their smartwatches (e.g. Sony, LG, AppleWatch, Fitbit Surge) or pedometers and wristbands (e.g. Fitbit Flex, Garmi n vivofit, Polar Loop, Jawbone).However, in the field of research, thither is a submit for more complex systems, which can provide more comprehensive information, of a larger variety.For this purpose, hybrid systems, combine the use of different techniques to compensate for the shortcomings of person systems. Such hybrid systems are equal by acoustic-inertial systems (Vlasic et al. 2007 Ward et al. 2005), e.g. conformation (Foxlin et al. 1998), optical-inertial systems, e.g. Hy-Bird (Ascension Technology) and inertial-magnetic systems, e.g. MERG sensors (Bachmann 2000), MTw development kit (Xsens Technologies), MVN Biomech and MVN Awinda (Xsens Technologies).Combined inertial and magnetic perceptual experience is soon one of the more popular choices in this area of study and go forth be discussed at length in the interest paragraphs.The light cant over, wireless and cheap, inertial sensors equipped with accelerometers, gyroscopes and magnetometers enable, when rateed on the homo body, the computation of angular orientation of the anatomical instalments to which they are committed to (Bellusci et al. 2013 Roetenberg et al. 2003).The on-board gyroscopes measure angular swiftness, based on the principle of angular here and nowum, according to the undermentioned fundamental comparison(1)Where crookedness on the gyroscope L angular momentum I moment of inertia angular velocity angular quickening.The most roughhewnly utilise gyroscopes for humanity motion studies are piezo-electric, capable of detecting vibration of majority.When an design vibrates while rotating, it is subject to the Coriolis Effect. This causes a second vibration to occur orthogonally to the sign vibration direction. The rate of turn can be calculate from this latter(prenominal) vibration. According to the pastime equations (2)Where m mass momentary speed of the mass with reference to the moving object to which it is attached.The resulting gyroscope signals are t hen delineate as being the sum of angular velocity t, offset due(p) to temperature of gyroscope bt, and white noise G,t (Eq. 3).(3)The gyroscope yield is very accurate, however, it is subject to errors and drift ca employ by integration of the signal over time, and the gyroscope temperature which can produce small offset errors, leading to large integration errors when reckon orientation.The use of compensatory estimation algorithms, such as Kalman filters can reduce the inwrought errors in the gyroscope output signal (Roetenberg et al. 2003). Kalman filters are mathematical algorithms use to efficiently minimize the mean of the squared error of a system output (Welch and Bishop 1995). Kalman filters are particularly useful for combining parameters of different cadence systems so that the advantages of one compensates for the weakness of the other, e.g. accelerometers are often used in conjunction with gyroscopes, in hostelry to compensate for inclination drifts in the gyros cope signal.The accelerometers measure the gravitational acceleration g and the vector sum of acceleration a. The output accelerometer signals are defined as the sum of acceleration at, sombreness gt and white noise A,t.(4)The inclination information provided by gt can be used to correct the orientation drifts of the gyroscope (Roetenberg et al. 2003). A further crude example of Kalman filtering, is using magnetometer readings to correct for the gyroscopes vertical axis drifts (Roetenberg et al. 2003).Magnetometers book the ability to detect local magnetic north and adjust psyche direction. The principles by which the magnetometers work are described by following equation(5)Where ym,t magnetic signals mt earth magnetic field vectordt haphazardness vectorvm,t -white noise.In real life measuring conditions the distribution of the magnetic field is often more complex and other parameters, such as changes in magnetic flux and the magnetic inclination angle, which can alter the magnitude of the magnetic disturbance, should be taken in consideration.The major limitations of using inertial and magnetic sensing for motion tracking are bringed by the following factorsFerromagnetic interferences can distort the local magnetic field and affect the measurements for the orientation about the vertical axis (Roetenberg et al. 2003).The velocity and type of cause performed and the geometry of the body segment to which the sensor is applied can affect the accuracy of the measurements (Roetenberg et al. 2005)Distances mingled with body segments cannot be assessed by essence of numerical integration (Roetenberg and Veltink 2005)Previous studies in which this type of equipment was used underwrite a high accuracy of the output data (Cutti et al. 2010 Ferrari et al. 2010a blind et al. 2014), however, the limitations in using this motion capture system are far from being overcome. The most important and challenging aspect of the study is to use the acquired informatio n in a biomechanically nitty-grittyful manner, e.g. the parameters declared as vocalise angles, bespeak to be as anatomically accurate as possible, for this purpose expect the juncture angles can be calculated as the angles of movement between two anatomical segments is not enough, a more complex mathematical model needs to be developed in order to squall the biomechanical characteristics of the studied correlative.There are a variety of protocols and algorithms available for maculation processing of sensor data stemming from human motion studies. A common approach for solving a human kinematics problem is to compare the human body to a robot manipulator. Similarly to a robot manipulator, which forms a kinematic chain from links interconnected by joints, the human body can be considered a kinematic chain formed of anatomical segments connected by articulations. In theory, this is a very efficient manner to solve a biomechanical problem.Cutti et al., for example, use the Dan avit-Hartenberg convention of robotics in their Outwalk protocol, which states that a kinematic chain with n joints get out have n+1 links (Fig 2.4). To solve the kinematics, a arrange system is rigidly attached to each link. In this case, when joint is actuated, the beside and its attached direct frame perform a motion. Whichever motion is performed by the kinematic chain, the orders of each point on are constant when expressed in the coordinate frame (Zatsiorsky 1998).The Danavit-Hartenberg convention has two conditions which need to be comfortable in order for the kinematic theme to be effective. The variables of a joint (e.g. gyration angles) are defined by the two coordinate systems of the links neighboring(a) to the joint. So, for example, the coordinates of the frame are expressed in the frame. Firstly, the orthonormality of the frames needs to be established, meaning needs to be perpendicular to . Secondly, the projection of in the frame ought to get acro ss .Comparing the human body to a robotics model is a well-grounded starting point. However, using the, frequently associated, strap-down integration method when measuring human kinematics with sensing units poses a very important limitation (Seel et al. 2014). The strap-down-integration method is based on using sensing units securely fixed to the even surfaces of robotic elements. However, thither is a significant difference between a robotic setup and an anatomical system.Firstly, aligning the sensor to an anatomical location, such that one axes of the sensor coordinate system coincides exactly with an axis of the anatomical joint, is nearly impossible (Seel et al. 2014). This rejoinder has been addressed in different manners by researchers so far.In the Outwalk protocol, Cutti and Ferrari et al. define as many coordinate frames for each link as the joints they form. Each anatomical segment has, therefore, a distal and a proximal coordinate frame. The joint variables are defined by the distal coordinate frame of one segment and the proximal coordinate frame of its adjacent segment.Another issue that needs to be addressed, when discussing a human biomechanical model, is an closely certain misalignment of the thigh axis with the segments coordinate system. Some studies completely cut the misalignment between the anatomical and the sensor axes (Seel et al. 2014). In the Outwalk protocol this problem is work out by expressing the flexure- elongation axis of the knee in the coordinate system of the distal femur and defining the other revolution axes of the coordinate frame as being orthogonal with respect to the new axis.This is another promising approach, however, in order for this method to be effective, the knee flexion-extension axis needs to be accurately place.In the case of hinge joints, such as the modify model of a knee joint, it is possible to calculate data from inertial sensors attached to both ends of the joint. However, this resulting data still needs to be translated into joint related coordinate systems and although, it is impossible to determine the initial position of the sensors on the anatomical segment, there is a possibility to determine the direction of the joint axes, by using different approaches to identify a functional movement axis from a set of dynamic motion data (Cutti et al. 2010 Ferrari et al. 2010a Seel et al. 2014).In their protocol Cutti and Ferrari et al. use Woltrings mathematical solution for determining the delimited helical axis (reviewed in (Zatsiorsky 1998)) to identify the knee flexion-extension axis. Woltrings solution appears to be fitting at least for most motion capture systems (Seel et al. 2014). However, the sensing units used in our study cannot measure translation. This would pose a rangy problem and could potentially result in substantial errors.In order for the outcome of the study to be successful, it needs to satisfy a set of conditions (1) it is very important that the resu lting post-processed sensor data is biomechanically meaningful to the musculoskeletal system (2) data acquisition needs to be user friendly, rapid and halcyon to complete (2) sensor mounting is not allowed to restrict the participants movement in any manner (3) the resulting data needs to relate to true anatomical joint angles and (4) the resulting information needs to be comparable to the reference system (Vicon).Seel et al. offer a solution based on rotational angle estimates alone, which is not only more simple from a data acquisition and processing point of view, but also functions on principles connatural to SARA and SCoRE.In the protocol proposed by Seel et al. the knee is assumed to be a simple hinge, with one sensor attached to each segment forming the joint. In order to compensate for the lack of information concerning the initial position of the sensors on the anatomical segments, the unit length direction vectors and the orientations of the two segments attached to the hinge joint (Fig 2.6) are estimated as described below.The Seel et al. solution only employs the use of what is considered to be raw accelerometer and gyroscope output data from the two sensors, the thigh sensor and the radical sensor. In reality, any output data produced by the Xsens sensors, used in Seel et al.s study and the menstruum study, is pre-processed in real-time by the on-board Kalman filter.For the purpose of the summary of the following protocol, all data indexed with 1 refers to thigh sensor data and data derived there from, and all data indexed with 2 refers to shank data and data derived there from.Firstly, the unit length direction vectors of the flexion-extension axis of the knee , are identified in the local coordinates of the sensors, by using an optimisation algorithm to inscribe the values of . Where the spherical coordinates for are(6)(7)With the following sum of squared errors (8)A search function is then used to find which satisfy the following conditi on(9)Where angular rates recorded by the thigh and shank sensor, respectively, with the sample period constant Euclidean norm.The acceleration measured by each sensor is the sum of the acceleration due to movement or so the joint centre and the acceleration due to the rotation of the sensor slightly the joint centre. In order to estimate the knee joint position expressed in the local coordinate systems of the sensors, the amounts by which are shifted in order to obtain the acceleration of the joint centre, are estimated first.Two compulsory points along the axes are estimated using a Gauss-Newton optimization algorithm. These points are shifted as close as possible to the sensor origin by applying(10)(11)The radial-ply tire and tangential acceleration due to the rotation of the sensor around the joint centre is computed i=1,2 (12)Where are time derivatives for angular rate and (13)The following sum of squared errors is calculated (14)A search function is used to find which satisfy the following constrain(15)The knee flexion/extension angle based on the gyroscope information is calculated with the following equation(16)The measured accelerations are shifted onto the joint axes by applying the following(17)(18)Where, represent the same quantity in the two different local coordinate systems, which rotate with respect to each other around the flexion axis.The flexion/extension angle calculated according to acceleration data can be defined as the angle between the projections of .(19)Where, and are pairs of joint plane axes, defined by The knee flexion/extension angle defined by fusing the accelerometer and gyro data is defined by(20)Where knee flexion extension angle calculated according to accelerometer data at time t knee flexion extension angle calculated according to gyroscope data at time t the weight of the accelerometer data.By using the most effective methods presented in the literature review, the current study will attempt to validate the inertial sensor protocol proposed by Seel et. al 2014 against a OSSCA method and to compare research lab and non-laboratory based inertial motion capture.