Heartbeat accuracy for biometric verification shows promise and limitations

A recent study published in Scientific Reports examines a new method of biometric verification called CompaRR, that records beat-to-beat information embedded in intervals between heartbeats.

While many forms of biometric verification involve stationary physiological signals like facial authentication and fingerprint scanning, non-stationary signals like heartbeats are more difficult to fake.

While ECG (or “EKG”) recordings have shown the capacity to verify identity through the heart’s electrical activity, the method uses a complex system that usually requires a medical professional to set up.

By using a method that measures beat-to-beat time intervals, wristbands and video cameras can gather the biometric data needed to verify identity in a more cost-effective, user-friendly way.

The study extracted recordings of heartbeats from 30 mice under anesthesia ranging from 6 to 24 months of age, which is equivalent to roughly 20 to 75 human years. Using the proposed method, researchers were able to verify the identities of the mice within fifty heartbeats, which is close to the resting heart rate for humans with a minimal equal error rate (EER) of 0.21.

For context, the accepted EER rate for most biometric verification methods ranges from 0.1 to 10 percent, the researchers say, depending on the modality and reference database size. ECG-based heart monitoring algorithms have been found to deliver an EER ranging between 0.2 and 19.15 percent. However, the researchers also noted limitations in the reliability of these accuracy evaluations.

The study found that EERs for mice were age-independent until the mice reached the age of 18 months, roughly equivalent to age 50 for humans. Verification was still possible when the model was trained on drug-treated mice, the study says. Verification was less accurate for mice that were awake, possibly due to differences in circadian rhythms throughout different states.

Researchers have previously found that the accuracy of heartbeat biometrics can be improved by considering signals typically associated with music.

Article Topics

accuracy  |  biometric identifiers  |  biometrics  |  heartbeat biometrics  |  identity verification