Validation results

Tested against the lab standard.

We tested BioAlign's walking analysis against laboratory motion-capture systems, and its fall detection on thousands of recorded falls and everyday movements. Here are the results.

Walking analysis Fall detection

Walking analysis

One ordinary camera, matching the lab on average.

We compared BioAlign with two laboratory systems: a 10-camera Vicon optical motion-capture system with healthy adults, and Xsens motion sensors worn by older adults. Every walking measure agreed with the lab on average.

−0.009 m/s

Average walking-speed difference from lab optical motion capture. A meaningful change in walking speed is 0.10 m/s.

0.011

Average gait-symmetry difference from lab motion sensors, measured on older adults.

4°

Average hip-angle difference from the lab system, on both legs.

6–11°

Average knee-angle difference from the lab system.

Healthy adults vs. Vicon optical motion capture

32 adults, 140 walking passes. Averages for all passes.

Measure	Lab	BioAlign	Difference
Walking speed	1.116 m/s	1.107 m/s	−0.009 m/s
Cadence	103.6 steps/min	98.9 steps/min	−4.7 steps/min
Stride length	1.433 m	1.341 m	−0.092 m
Gait symmetry (0–1)	0.971	0.977	+0.006

Older adults vs. Xsens motion sensors

11 older adults with complete records.

Measure	Average difference	Average size of difference
Gait symmetry	+0.005	0.011
Cadence	−6.4 steps/min	10.9 steps/min

A single walk varies, and the lab's own readings vary from walk to walk too. BioAlign builds each person's picture from many walks over time. In this test, averaging each person's walks cut the spread in half (walking speed ±0.42 m/s for one walk, ±0.20 m/s per person).

Fall detection

Catches falls, with few false alarms.

We tested on 4,827 recorded clips from nine public research datasets: 2,261 falls and 2,566 everyday activities like sitting down fast, lying on a bed and bending over. Every clip ran through the same software BioAlign runs in real installations.

87.7%

of falls caught at a facility the system had never seen

8.6% false alarms

94.3%

of falls caught at a facility like the ones it learned from

6.3% false alarms

93–97%

of falls caught on the two largest datasets, three-quarters of all clips, even when unseen

1–2% false alarms

The "never seen" result is the honest one for a new installation: the system was tested on each dataset after learning only from the other eight. Most published fall detectors only report the easier familiar-facility number.

What's next

From research recordings to real rooms.

These results come from public research recordings, not a live facility. On-site testing in homes and senior living communities is the next step, and we'll post those results here.

Talk with us See the platform

Studies by Don Kolz, KinovaAI LLC. Walking data: GPJATK dataset (Vicon optical motion capture) and the Toronto Older Adults Gait Archive (Xsens inertial motion capture). Fall data: FallVision, VFD29, OCCU, EDF, CAUCAFall, GMDCSA-24, Le2i, UR Fall and UMAFall. The author has a financial interest in KinovaAI LLC. BioAlign Care is a wellness and safety monitoring aid, not a medical device or a diagnostic tool. © KinovaAI LLC, 295 W Cromwell Ave #108, Fresno, CA 93711.