Synchronization Of Video Kinematic And Analog Biomechanical Data Using The Motion Analysis System

Abstract

A significant advantage of using integrated biomechanical data collection systems to study human movement is the simultaneous storage in a single computer of video-derived kinematic data and digitized records of analog data such as force and electromyographic signals. However care must be taken to assure that these records are properly synchronized by the respective data-logging parts of the hardware and software. This is particularly important when the data are combined to calculate dynamic values such as net joint moments. We have studied data synchronization using the Motion Analysis Corporation's Expertvision 60Hz video system (VP320) and have identified three important issues: 1) sampling rate; 2) video sampling initiation; and 3) intra-frame video/analog initiation synchronization. The first issue, sampling rate, arises since each data collection process assigns time values throughout the data based on a nominal sampling rate. Errors in this assumed rate will cause increasing cumulative synchronization error throughout the sampling interval unless appropriate corrections are appplied. We have found the analog data collection rates to be sufficiently accurate (within the lOO-2KHz range), but the video sampling rate deviates from the nominal 60Hz by 1-2.5%, depending on the hardware version. This error requires a software adjustment to the packaged calculation of video frame times. The second issue, video sampling initiation, occurs because the video collection triggering process starts by storing (as "Frame 1") the second video field to occur after the detection of the trigger signal. Since the A/D conversion of analog signals uses the same algorithm but (usually) a much faster sample rate, the analog sample begins nearly one video frame interval ahead of the video sample. This constant initiation time offset must be corrected before post-hoc data analysis. The third issue compounds the initiation correction procedure, as the trigger signal (and subsequent analog data collection) may occur at any point between two video frames, introducing an additional random offset ranging from 0-1 7ms. In many applications this amount of error is unacceptable, and correction in the form of external video synchronization is required. One simple way to accomplish this is to lead the video synch pulse into an unused A/D channel; another is to use a pulsed light in the video record which is also sampled by the A/D system. Synchronization errors are system-specific, and may be constant or variable. Collection of accurately synchronized video kinematic and analog data requires careful attention and post processing to correct these three sources of potential error.