State-of-the-art biomechanical analysis of human motion requires specialized equipment operated within a laboratory. In a typical experiment, a subject is asked to run across a room multiple times with reflective markers attached to his/her skin, while motions are recorded on specialized video equipment for later computer analysis. This equipment, however, has limitations. Video cameras capture only the reflective markers, not the human body. These markers are tedious and time-consuming to apply, feel awkward to the subject, and limit experiments to laboratory-controlled environments.
Recently, a motion-analysis system has been developed at Stanford University (Stanford, CA, USA; www.stanford.edu) that uses high-speed cameras tied directly to multiple PCs through FireWire cables. Instead of identifying markers, system software identifies entire body parts based on patterned clothing rather than on reflective markers and specialized lighting.
Structured light and four pairs of stereo cameras determine the 3-D surface shape of a leg. Synchronizing the cameras allows more accurate surfaces to be generated since all of the pictures are taken at exactly the same time. If the subject moves and the cameras are not synchronized, the surface maps do not match.
The biomotion imaging system uses image-processing software from NorPix (Montreal, Quebec, Canada; www.norpix.com) and is based on eight Basler (Exton, PA, USA; www.baslerweb.com) 8301 FC monochrome cameras. The eight cameras form four stereoscopic camera pairs around a runner: front, back, left, and right.
"For running, people use 120 frames/s or 240 frames/s. For walking, 30 frames/s is too slow, so 60 frames/s is normally used. Basler's 80 frames/s was a good trade-off between speed and price. Other systems were faster but were more expensive. Previously, we used consumer video recorders for this, so the new synchronized camera system is a big improvement." says Ajit Chaudhari, a biomechanical engineer at Stanford.
Each stereoscopic pair feeds into its own FireWire card in a PC desktop with 2.6-GHz Pentium 4 running Windows XP. Out of the four PCs, which are networked using standard Ethernet connections, one PC is designated as a master. This PC contains a 6601 dual signal-generator card from National Instruments (Austin, TX, USA; www.ni.com). LVDS cables run from the 6601 card to each of the cameras providing a single trigger/synchronization signal for all the cameras and PCs. The 6601 triggers all the cameras simultaneously while providing timing information so that each PC can time stamp each frame for later biomechanical motion analysis using Chaudhari's proprietary image processing algorithms.
Each PC also runs two instances of NorPix's StreamPix software—one for each camera. StreamPix displays the video and time-stamps for each frame before saving the streaming video to the hard drive in a variety of compressed or uncompressed formats. The master PC also contains NorPix's StreamNetServer software, which allows the operator to control all eight instances of StreamPix from a central GUI.
"Although multiple PCs are required in the system, the elimination of conventional frame grabbers reduces system costs. With specialized cameras, expensive cables may be required, and the system cannot be changed easily. With FireWire, hubs and cables can be purchased from consumer stores and the system easily reconfigured," Chaudhari said.