Detection of Human Blood Pulse Based on Displacement Vector in Video Footage

This paper presents a simple yet effective method of blood pulse rate estimation based on video footage processing. The 15-s scenes are acquired from the skin over big arteries with a regular smartphone camera and subtle skin displacements are detected and analyzed based on individual irregularities of skin texture. The rhythmic component is then distilled from displacement information and qualified as representative of blood pulse. The method was evaluated in two experiments: the first aiming at covering all frequency range 55-140 bpm typical to rest and mild exercise at home, and the second aiming at measurements in 15 different people. The proposed method shows systematic error of 0 bpm ±1.86% and -0.8 bpm ±3.0% respectively against the reference measurement with a pulse oximeter. This result was found satisfactory for home-based self-monitoring to pursuit the circulatory response to exercise, stress and medications. Moreover the method is hygienic, may be applied to various parts of the body and does not require dedicated equipment.