Analog-to-Digital Converters Analog-to-Digital Converters (ADC) convert voltage to a digital value that represents the signal at the instant at which it was sampled. ADC integrate third-party devices for capturing analog data into a Vicon system. Analogue devices ADC Vicon MX controller 1 Signal type Bipolar: (i.e. ±5.0V), full range output 10V Unipolar: (i.e. between 0-10V) Dynamic Range Analog voltage signals digitized over a particular range Input range • i.e. min and max input voltage is defined over which the quantization should occur. • bipolar, i.e. range ±5V (full range 10V) • unipolar, range from 0 to 10V Many A-D converters have an actual range of 0 to 10 V, so an input signal with a range of 0 to 50 mV needs to be amplified by a gain of 200 before it can be converted. Choose range that closely encompasses your signals, as this will increase the resolution. 2 When sampling the voltage signal with an ADC, two variables must be considered: 1. Sampling precision (resolution) How many different gradations (quantization levels) are possible when taking the sample. 2. Sampling rate Samples per second Resolution (quantization) ADC have fixed number digits, described by binary digits, or “bits” available for quantifying the voltage signal detected at the input. ADC resolution as a function of bit length Number of Bits Range of n-bit ADC = 2n bits • i.e. 12 bits describe 4096 values Unique Values 2 4 4 16 8 256 12 4,096 14 16,384 16 65,536 ADC Resolution Based on the number of steps the input range1 is divided Many ADC’s have full range of 10 V (±5V), if an input signal has a range of 50 mV, it needs to be amplified2 by a gain of 200 before it can be converted. 1 Max and min voltage that will be digitized. increases the voltage signal to a level suitable for the ADC 2 Amplification 3 Resolution is expressed as Vresolution = Vrange / (2n) For a 4 bit ADC with a range of ±10 Volts, (full range 20V) Vresolution = 20V/(24) =1.25 V =1250mV • Resolution for a ±10V range will be resolved to 1250mV, • Resolution for a ±5V range will be resolved to 625mV. Resolution increases when the input range is narrowed, Steps from 0 to 15 scaled represent full input voltage range (i.e. ±5 V) Bits 0 to 7 represent the negative voltage range Bits from 8 to 15 represent the positive voltage range. ±5.0 V ADC with 16-bit resolution • 65,536 discreet samples mapped between 0 - 65,536 bits • OFFSET is 32,767 for a symmetrical -5 to +5 V range • equally mapped directly as signed integers in the range of –32,767 to + 32,767 OFFSET would be 0. • If ANALOG:SCALE and OFFSET applied correctly, both configurations return identical values for the -5 to +5V range. 4 Typical input ranges for ADCs are ±5 and ±10 Volts. • 16- bit ADC with an input range of ±5V has a precision of 153 μV. • ADC specified for a range at ±10V would have a precision of 305 μV. Note: no point in trying to resolve signals below the noise level of the system: all you will get is unstable readings. It is important to ensure that the ADC range encompasses the full span of the voltage input while maintaining the minimum resolution necessary. Nyquist theorem Sampling frequency should be at least 2X the highest frequency contained in the signal. fs ≥ 2fc fs: sampling frequency (samples per unit of time) fc: highest frequency contained in the signal. Sine wave at a frequency of 1 Hz Sampling @ 2 Hz capture each peak > 2X captures more variation in the signal 5 Aliasing (sampling errors) < 2X sampling causes highfrequency components to be aliased with genuine lowfrequency ones, resulting in incorrectly reconstructed waveforms during DAC. Complex signals contain many frequency components. • Fourier's theorem, continuous signals may be decomposed in terms of a sum of sines and cosines at different frequencies. • 1 Hz, 2 Hz, and 3 Hz sine wave added together. Nyquist theorem is 2X the highest frequency, irrespective of how many other frequency components. fc = 3 Hz, fs ≥ 2fc, ∴ fs ≥ 6 Hz Digital representation of the original signal By increasing both the sampling rate and the precision, you reduce sampling error . both the rate and the precision have been improved by a factor of 2 the rate and the precision have been doubled again 6 Motion Capture Systems & Analysis Motion studies date back to the 19th century. • Eadweard Muybridge (1878) most widely known. Stanford and the gallop question • • • • Successive-exposure photography 50 cameras along a track parallel to the horse Shutters triggered by trip wire by the horse's hooves. Hooves leave the ground at the moment when all the hooves are tucked under the horse. Developed methods to capture human motion • Pioneer in the Science of Biomechanics and the mechanics of human movement. Biomechanics Yellow Pages http://isbweb.org/c/isb/pub/files/orig_website/~byp/Category_List.html Animazoo Intelitek Qualisys AB Ariel Dynamics, Inc. JC Labs Redlake Biogesta Kine ehf. Robotic Systems Ltd Biomechanical Solutions Mechanical Dynamics Inc. Silicon COACH Ltd BTS SpA - Bioengineering Medical Imaging Systems Charnwood Dynamics Ltd Motion Imaging Corporation Skill Technologies, Inc. Computerized Function Testing Corporation Mikromak Spica Technology Corporation, Inc. CONTEMPLAS GmbH Motion Analysis Corporation Sport and Physical Education Technology Ltd. Darras Software Development NAC Image Technology STT Engineering and Systems Digital motion analysis Northern Digital True Tape, LLC. Electronic Imaging Systems OsteoKinetics Corporation Vaquita Software eMotion Peak Performance Technologies, Inc. Vicon Motion Systems SIMI Reality Motion Systems GmbH GaitMat II PhoeniX Technologies Webbsoft Solutions Image Diagnostics Ltd Photo-Sonics welch-e technologies Innovative Sports Training, Inc. Photron Zebris GmbH Innovision Systems, Inc. Pixoft Diagnostics Imaging Ltd zFlo, Inc. Electrogoniometers Biometrics Ltd. Inexpensive & approximate measure of joint angle. Placed across the joint, One or two or potientimeters between two bars. • Uniaxial: rotations in one plane • Biaxial: simultaneous measure of two rotations Potentiometer produces change in electrical output depending on angle Data logger accommodates EMG, electrogoniometers, accelerometers. 7 Advantages • portability for collection in the workplace or other sites • ease of set up and processing • relatively low cost • permit collection and storage of large data sets over a prolonged period Disadvantages • lack of data with respect to the global reference system and 6-DOF • errors due to alignment of the axes of rotation • difficulty in monitoring joints surrounded by large amounts of soft tissue cross talk between potentiometers. Electromagnetic tracking systems Flock of Birds, Polhemus Transmitter • (low-frequency magnetic coils) emit an electromagnetic field. • reference frame for sensor measurement. Receiver • small lightweight cube • enclosed are electromagnetic coils that detect magnetic fields. • position & orientation are precisely measured as it is moved. • are completely passive devices, having no voltage applied to it. Advantage • Elimination of marker dropout from camera FOV (which can occur in videography), • real time 6-DOF data, • Accuracy Limitations • • • • Cable connecting sensors can inhibit movement, sensors slippage, Limited number of sensors can be tracked at one time and cost. Interference from metallic objects or other magnetic fields degrade performance. • distortion effects normally seen with long range electromagnetic systems. • Cost 8 Optoelectronic systems OptoTrak, CODA,Selspot Active markers • infrared emitting diodes (LED’s) placed on the segments or joints. • LED’s triggered and pulsed sequentially by a computer, permitting automatic identification of each marker. • Position Sensor detects horizontal and vertical movement of each active marker. • Independent marker frequencies enables you to capture data at even higher speeds. Advantages • automated marker tracking, no marker merging or misidentification. Limitations • • • • • LED’s require wire connection, Cost Measurement cumbersome Limited to the laboratory More than one camera bank may be required to obtain adequate marker coverage. Videography APAS, ElitePlus, Motion Analysis, ProReflex, Vicon Most frequently used. 1 or more cameras track passive reflective markers. • Passive markers reflect either external ambient light or camera-projected light (infrared). • markers then reflect the light back into the camera lens, and the digital signal is fed into a computer. A threshold is set to automatically discriminate the marker “pixels” which are the brightest objects in the laboratory. Track horizontal and vertical coordinates of each marker from each camera. 9 2D systems • one camera • Assumption: all motion is occurring perpendicular to the camera axis. • Seldom, marker movement outside this plane will be distorted. • 2D systems are less accurate than 3-D systems 3D systems • 3D coordinates computed using 2D data and the known location from each camera using the principle of optical triangulation. • Many systems now employ >6 Advantages • • • • Track large numbers of markers High sampling rates (>1000Hz) the potential for high precision and accuracy Unencumbered movement Limitations • Potential marker merge from various camera views. • Limit on how close markers may be placed (> 2mm) • Marker label loss resulting from marker occlusion, dropout or trajectories crossing – Missing points filled by interpolation. • Loss of resolution with high sampling rates Vicon Camera performs 2 functions; • Illuminates capture volume with infrared strobe light. • Reflections from surface markers seen by each camera. Pixel resolution and sampling inversely related • Higher resolution, the lower your sampling rate. • 2352x1728 pixel images up to 160 Hz 10 Marker centroid derived using edge detection or 10-bit grayscale precision. More pixel resolution from the markers, greater accuracy calculating relative 2D marker position in the 2D FOV. Vicon: Sampling rate and resolution MX40 full resolution is over 4 million pixels (2352 H x 1728 V) ≥ 160Hz, decreased camera resolution reduce (vertical windowing). ≥ 500Hz, cameras resolution is 2MP. Max sampling rate up to 1000Hz. Turn cameras on side for greater vertical resolution during high speed capture. 11 Purpose of Calibration Assist in reconstructing 3D positions of captured markers from 2D camera images • To establish an absolute reference of origin and orientation Earlier calibration and reconstruction methods • Direct Linear Transformation (DLT)Abdel-Aziz YI and Karara HM (1971) DLT 3-D algorithm contains up to 22 parameters. to determine all parameters for reconstruction procedures, calibration devices consisting of precisely measured marker points are used to determine any unknowns. • Non-Linear Transformation (NLT) Dapena J, Harman EA, and Miller JA (1982) Direct Linear Transformation (DLT) Cameras calibrated using control points fixed to a rigid calibration frame. Accuracy determined by the precision of the coordinates for the control points and digitizing errors. Measurement volume limited by the size of the calibration frame • accuracy increased as control points increased • Best accuracy achieved with even distribution • Accuracy decreased as distance from control region increased Chen 1994 Static & Dynamic Calibration y Static calibration Establish absolute spatial reference Utilizes L-frame with two axes • • • • x Establishes origin of lab space Combines with the first axis to establish the 2nd axis Resultant cross product is the third axis Origin specified in CRO file Dynamic calibration • Determines relative camera position & orientation • Linearization – optical distortion from camera lenses measured – correction matrix calculated – corrections applied to each camera for every frame during MoCap 12 Algorithm uses known distance between markers to establish the scale of the measurement volume. Additional cameras initiate optimization algorithm to determine best fit. This distance constant is specified in the Calibration Reference Object (CRO) file. Reconstruction Mathematical transformation of 2D images from each camera into a 3-D coordinate system. Based on precise knowledge of both internal (lens and camera characteristics) and external (spatial orientation) camera parameters. Camera residual A measure of the accuracy of one camera relative to all other cameras Derived from reconstructed wand marker positions from all cameras. • Camera in question does not contribute to its own 3D reconstruction residual The distance between the reconstructed markers image and the camera’s own image averaged across all available samples – High camera residual = 2D contribution that tends to be less accurate than the other cameras 13 Mean Residual – average of individual residuals Residual Range – the highest and lowest residuals Wand Visibility – the average percentage of the wand wave that contributed to each camera’s calibration Static Reproducibility – a measure of the accuracy of the reconstructed positions of the L-frame markers compared to the CRO file Common reasons for failed calibrations Inappropriate wand wave • Too fast (low frequency) • Solution: slow wand wave Poor camera positioning • Inadequate overlap • Solution: – adjust cameras – May need to customize wand wave Background static noise • Reposition cameras • Adjust camera sensitivities (can be turned up after calibration) Wand wave tips: • Avoid breaking cameras’ views with one’s body. • Position body so the wand can be seen by as many cameras as possible. • Move around the entire volume to give all cameras an equal opportunity to see the wand wave. • No need to collect more than 1000 samples 14 Definitions Residuals –indicate the quality of calibration Linearization –process of correcting for lens distortion Bundle Adjustment – an algorithm that performs the linearization process by optimizing the camera parameters to give the best re-projection error • In signal processing, • sampling is the reduction of a continuous signal to a discrete signal. i.e. conversion of a sound wave (a continuous-time signal) to a sequence of samples (a discrete-time signal). • Exact reconstruction of a continuous-time baseband signal from its samples is possible if the signal is bandlimited and the sampling frequency is greater than twice the signal bandwidth. 15