Body Weight Estimation Using Smartphone Based Photoplethysmography Signal

This paper presents a novel approach for automatic estimation of human weight using only the smartphone's built-in camera-based fingertip video. PPG signals are analyzed for weight prediction utilizing their correlation with physiological characteristics such as blood volume, heart rate, and respiratory rate, which are intimately related to body weight, as evidenced by individuals with greater body mass having higher blood volume, heart rates, and respiration rates. In this study, 92 consenting participants provided fingertip videos during the data collection phase, placing their fingers on the smartphone's camera lens, and using the camera flashlight for illumination. Then, each subject's fingertip video was processed to provide PPG signals at a spectacular frame rate of 60 frames per second (fps). Advanced preprocessing techniques were used to successfully eliminate noise and baseline drift in the PPG signals captured at 60 fps, enabling precision tracking for precise weight measurements. A DNN model is trained to evaluate the measurement using the 40 different features that were extracted from the PPG signal using Fourier analysis, first and second order derivatives, and single period. The potential for this innovative approach to change health and wellness applications is demonstrated by the estimated R
²
= 0.96 and MAE=2.642 kg which is comparable to commercial weight scales. Due to the widespread use of smartphones, this technology transforms weight measuring by offering an accessible alternative to conventional procedures that rely on specialized weight measurement devices, which are often out of reach for many individuals.